Color Matching and Color Discrimination
Uploaded by
leiansantos21Color Matching and Color Discrimination
Uploaded by
leiansantos21See discussions, stats, and author profiles for this publication at: https://www.researchgate.
net/publication/254694907
Color Matching and Color Discrimination
Article · July 2003
DOI: 10.1016/B978-044451251-2/50004-0
CITATIONS READS
64 677
2 authors, including:
Joel Pokorny
University of Chicago
295 PUBLICATIONS 12,735 CITATIONS
SEE PROFILE
All content following this page was uploaded by Joel Pokorny on 29 September 2023.
The user has requested enhancement of the downloaded file.
3 Color Matching and
Color Discrimination
Vivianne C. Smith and Joel Pokorny
Departments of Ophthalmology &Visual Science and Psychology,
University of Chicago, 940 East Fifty-Seventh Street, Chicago, IL
60637, USA
CHAPTER CONTENTS
3.1 Introduction 104 3.3.3.1 Displacement laws 129
3.2 Color mixture 104 3.3.4 Increment detection on white 132
3.2.1 Principles and procedures 104 3.4 Chromatic discrimination 132
3.2.2 Representation of color matching data 105 3.4.1 Historical approaches 133
3.2.3 CIE standard colorimetric observers 110 3.4.2 Experimental variables 135
3.2.4 Experimental variables 116 3.4.3 Modern approach to chromaticity
3.2.5 Interpretation of color matching 117 discrimination 136
3.2.6 Sources of individual differences in color 3.4.4 The effect of surrounds 137
matching 120 3.4.5 Interpretation 137
3.3 Chromatic detection 124 3.5 Congenital color defect 138
3.3.1 Threshold-versus-radiance (TVR) functions 125 3.5.1 The protan and deutan defects 138
3.3.2 Explanatory concepts in detection 125 3.5.2 Tritan defects 141
3.3.2.1 Adaptation 125 3.6 Acknowledgment 142
3.3.2.2 Saturation 127
3.3.2.3 Noise 128 3.7 Notes 142
3.3.3 Detection on spectral backgrounds 128 3.8 References 142
The Science of Color Copyright © 2003 Elsevier Ltd
ISBN 0–444–512–519 All rights of reproduction in any form reserved 103
■ THE SCIENCE OF COLOR
3.1 INTRODUCTION neural state in which neural signals generated by
the fields are identical.
The purpose of this chapter is to summarize the
data of color matching, detection of spectral 3.2.1 PRINCIPLES AND PROCEDURES
lights, and discrimination of lights on the basis of
differences in chromaticity. Matching, detection, Metamers: Metameric lights are lights that
and discrimination are all tasks in which the though of dissimilar spectral radiation are seen
spectral content of the lights is an important as the same by the observer. In a prototypical
parameter. The tasks involve either identity color-matching experiment using additive lights,
matching or detection of a just noticeable differ- the metamers are presented in a bipartite field.
ence. Data are expressed in physical units. The For 2 foveal fields, metamers have three impor-
color appearance at the identity match or at the tant properties that allow treatment of color
just noticeable difference is irrelevant and rarely mixture as a linear system (Grassmann, 1853):
noted. Although color names are often used to
1. The additive property. When a radiation is
refer to lights as a shortcut form of expression
identically added to both sides of a color
(e.g. red light for a 660 nm spectral light or white
mixture field, the metamerism is unchanged.
light for a continuous spectral power distribu-
2. The scalar property. When both sides of the
tion), we have tried to avoid such terminology in
color mixture field are changed in radiance
this chapter.
by the same proportion, the metamerism is
The data of matching, detection, and discrimi-
unchanged.
nation have been modeled successfully by con-
3. The associative property. A metameric mix-
sidering early retinal processing. All the
ture may be substituted for a light without
phenomena described in this chapter can be
changing the metameric property of the color
modeled by the activity of parvocellular (PC-),
fields.
koniocellular (KC-) and magnocellular (MC-)
pathways (see Chapter 6). Some readers may According to Grassmann’s laws, a color match is
find it helpful to read Chapter 6 concurrently invariant under a variety of experimental condi-
with this chapter. tions that may alter the appearance of the
We will concentrate primarily on the classical matching fields. Metameric matches will hold
studies in the literature. These studies have with the addition of a chromatic surround or fol-
involved small (1 or 2) or moderate (10) sized lowing pre-exposure to a moderately bright
foveally fixated, circular stimuli. Luminance lev- chromatic field.
els of matching and discrimination stimuli are
usually at low to medium photopic levels, with Trichromacy: A fundamental property of nor-
retinal illuminations of 5–1000 trolands (td). In mal human color vision is the existence of color
detection studies, large concentric backgrounds matches of lights that differ in spectral composi-
are often employed and light levels may extend tion. It is possible to find a metamer for any light
to levels where substantial photopigment (spectral power distribution) by variation of the
bleaching occurs. energies of three fixed lights, which are called
primaries. The terms trichromat and trichromacy
refer to this property of human color vision.
3.2 COLOR MIXTURE There is wide freedom in the choice of primaries.
A formal requirement is that one primary cannot
A human observer looking at a colored object be metameric to a mixture of the other two. In
has no way of knowing from its appearance the practice, it is desirable that the primaries be spec-
spectral composition of the physical stimulus. trally separated as much as possible and that the
Colorimetry provides a system of color measure- matching field is at a mid-photopic level. The
ment and specification based upon the concept choice of primaries is dictated largely by experi-
of equivalent-appearing stimuli. In a color mental convenience. Primaries may or may not
match, two fields of differing spectral radiation themselves be spectral. However, in the develop-
appear identical. The match represents a unique ment of a colorimetric system, the test lights
104
COLOR MATCHING AND COLOR DISCRIMINATION ■
should be spectral or near-spectral in order to (1947) and Le Grand (1968). The data on which
derive the largest color gamut. 2 colorimetry was formulated were those of
It has become customary to present the results Wright (1929, 1946) and Guild (1931). Stiles
of color mixture experiments as color equations. (1955) has discussed this work and presented
Suppose we have three primaries, P1, P2, P3 and modern data for both 2 and 10 fields. Here we
a test light, S, arranged in a bipartite field. We describe Wright’s experiment and how the
find that a mixture of S and P3 appears identical colorimetric observer was formulated.
to a mixture of P1 and P2, when the radiant Wright used a technique known as maximum
energies of S, P1, P2, P3 are PS, PS,1, PS,2 and PS,3 saturation matching. The primaries and the spec-
respectively. This is written: tral test lights are arranged pairwise in a bipartite
field so that a mixture of the spectral test light
PSS PS,3P3 PS,1P1 PS,2P2 (3.1) and one primary are matched to a mixture of the
two other primaries. The spectral primaries were
where ‘’ means visually identical and ‘’ at 650 nm (P1), 530 nm (P2), and 460 nm (P3).
means physical mixture. The quantities PS,1, PS,2 The spectral test wavelengths varied in 10 or
and PS,3 are called the tristimulus values. Their 20 nm steps between 410 and 700 nm. For each
subscripts identify the test light (S) and the pri- test wavelength, a color match was obtained and
maries (1, 2 or 3). Using Grassmann’s law we the amount of each primary was measured and
treat this mixture equation as an algebraic equa- expressed as an equation. Every spectral wave-
tion and express the match in terms only of S: length could be specified in terms of the three
primaries, with one of the three being negative
PSS PS,1P1 PS,2P2 PS,3P3 (3.2) or zero. Wright also made a match to his instru-
ment broadband (white) source, which was fil-
where the minus sign reflects the fact that in the tered tungsten with a color temperature of
color match, the primary P3 actually was added 4800 K. This source was essentially equivalent to
to the test light. Given a set of primaries, (P1, P2, one adopted formally by the CIE and termed
P3), a color match can be made to all lights, of Standard Illuminant B.1
any spectral power distribution. When the test
light is a narrow spectral band, one of the pri- 3.2.2 REPRESENTATION OF COLOR
maries is always either negative; i.e. physically MATCHING DATA
superimposed with the spectral test light to match
the remaining two primaries, or zero. When the The quantities of the test wavelengths are cus-
test light has a broad spectral power distribution, tomarily referred to an equal energy spectrum.
the three primaries may all be positive. There are several conventions to represent the
As a consequence of Grassmann’s law, the unit of measurement for the primaries (Wright,
match to a broad spectral power distribution can 1946; Stiles, 1955; Stiles and Burch, 1959; Le
be considered as the sum of constituent narrow- Grand, 1968). It is convenient for the purpose of
band spectral matches. Consider a distribution of discussion to redefine the tristimulus values PS,1,
unit energy at every spectral wavelength (the PS,2, and PS,3 of equations 3.1–3.2, using new
equal energy spectrum, abbreviated EES). terms, CS,1, CS,2, and CS,3:
According to the law of additivity, the color
match to this distribution is considered as the CS,i PS,i/ei (3.4)
sum of the matches to unit energies of the
spectral wavelengths. where PS,i is the radiant flux of primary (i) spec-
ified relative to a radiant unit, ei. If the three
PEESEES [(Pk)] EES [(Pk,1)] P1 radiant units are equivalent and unity, the data
(3.3)
[(Pk,2)] P2 [(Pk,3)] P3 will be expressed in watts, as implied in equa-
tions 3.1–3.3. In practice, however, other nor-
malizations are used. In energy units, the
Data of color matching: Early measurements short-wavelength C1 primary amount is much
of color matching are reviewed by Bouma higher than the others. In representing color
105
■ THE SCIENCE OF COLOR
mixture data, it was considered desirable that
P2
the amounts of the primaries have similar scales.
Recalculating a new normalization simply re-
quires weighting all the values of Ci. Suppose e’i
is the new primary unit: 0.8
PS,i CS,i e’i CS,i ei (3.5)
0.6
CS,I CS,i ei/ e’i (3.6)
0.4
Chromaticity diagrams: The data of a color B
matching experiment can be expressed in terms
of vectors in a three-dimensional space 0.2
(Schrödinger, 1920). The tristimulus values, C1,1, 0.6 0.8
0.2 0.4
C2,2, and C3,3, at the three primaries form unit P1
vectors. A resultant vector represents the location
of a mixture of the three primaries. According to 0.2
0.4 0.6 P3
the Grassmann laws, the scaling of the unit vec- 0.8
tors is arbitrary; if the radiance were doubled,
the entire diagram would be expanded but
would maintain its shape. A two-dimensional Figure 3.1 Representation of Wright’s color
unit plane can be defined in the space where the matching data in a three-dimensional space.The three
primaries P1, P2, P3 form the axes. Wright’s
sum of the tristimulus values is unity. The rela- chromaticity coordinates for the spectral test lights
tive positions of the resultant vectors are pre- are shown plotted in the unit plane.The arrowhead
served in such a plane. The tristimulus values are represents the vector for source B.
converted into a form where the sum of the
three always equals unity. For a test color S:
trum locus. Standard Illuminant B appears near
cS,1 CS,1/[CS,1 CS,2 CS,3], (3.7) the center of the right-angled triangle as point
cS,2 CS,2/[CS,1 CS,2 CS,3] (3.8) SB.
cS,3 CS,3/[CS,1 CS,2 CS,3]. (3.9)
WDW normalization: W.D. Wright performed
The quantities cS,i are termed chromaticity his experiment using a method of normalization
coordinates (trichromatic coefficients in the now termed WDW normalization. This normal-
older literature). The unit plane for Wright’s data ization is a practical choice, because it ensures
is plotted in Cartesian coordinates in Figure 3.1. that the tristimulus values will have rather simi-
The horseshoe shape shows the location of the lar weights, but does not require the experi-
spectral test wavelengths and the head of vector menter to have an absolute calibration of the
B represents the location of Wright’s instrument radiant energies of the primaries and test lights.
white (Standard Illuminant B). There is additionally a theoretical significance to
The chromaticity diagram is a two-dimensional data expressed in this normalization.
representation of the results of a color mixture In WDW normalization, two test wavelengths,
experiment. Figure 3.2 shows Wright’s data the first intermediate between P1 and P2 and the
replotted in a chromaticity diagram where c1 is second intermediate between P2 and P3, are cho-
plotted against c2 on Cartesian axes. The co- sen as normalizing wavelengths. Wright used
ordinates for P1, P2, and P3 form a right-angled 582.5 nm and 494 nm. The energy units are
triangle. Points that fall outside this triangle chosen so that the amount C1 of the P1 primary
represent the negative values of the spectral tris- is set equal to the amount C2 of the P2 primary
timulus values. The values for the spectral wave- at the match to the first normalizing wavelength
lengths again demonstrate a characteristic and the amount C3 of the P3 primary is set equal
horseshoe-shaped curve; this is called the spec- to the amount C2 of the P2 primary at the match
106
COLOR MATCHING AND COLOR DISCRIMINATION ■
520 530
510 P2 540
550
C2
560
500
570
Chromaticity Coordinate, c2
580
0.5 F
E
490 SB 590
600
D
610
480
620
630
470
P3 0.5 C1 650 P1
460
450
440
Chromaticity Coordinate, c1
Figure 3.2 The chromaticity diagram calculated for Wright’s data with WDW normalization and C2 plotted
versus C1.The coordinates for SB represent the position of the average match to Standard Illuminant B made by
36 observers, at coordinates (0.243, 0.410).The triangle DEF shows the spread of the individual white matches.
(From Y. Le Grand, Light, Colour, and Vision, translated by R.W.G. Hunt, J.W.T.Walsh, and F.R.W. Hunt, Dover
Publications, New York, 1957; reprinted with permission.)
to the second normalizing wavelength. In Figure tics of the visual photoreceptors. Individual vari-
3.2, the spectrum locus is the averaged data of 10 ation in the distribution of the coefficients for
observers. The coordinates for SB represent the the white point greater than experimental error
position of the average match to the instrument must be attributed to variation in prereceptoral
white made by 36 observers, at coordinates filters, e.g. the lens and macular pigment
(0.243, 0.410). The triangle DEF shows the (Wright, 1946; Wyszecki and Stiles, 1982) (see
spread of the individual white matches. Figure Chapter 2).
3.3 shows the chromaticity coordinates for the
spectrum for Wright’s observers plotted as a Photometry and colorimetry: The discussion
function of wavelength. The solid line shows the until now has assumed that the primary units
average data for 10 observers and the dashed are specified in radiant energy. However, that is
lines show the matches of extreme observers. not how colorimetry developed. Historically, the
The interobserver variability is partitioned radiant energy levels were not available in the
among the spectral wavelengths and the white early studies of Wright and Guild. Wright cir-
point. cumvented this problem by use of the WDW
The importance of the WDW normalization is normalization and he expressed his data in
that it separates interobserver variance caused chromaticity coordinates.
by receptor variation from interobserver vari- Equation (3.4) can also be expressed in lumi-
ance caused by prereceptoral variation. It may be nous quantities, F:
shown algebraically that in the WDW system,
interobserver variation in the chromaticity coor- Fkl KmVklPk (3.10)
dinates for the spectrum greater than experi-
mental error can be attributed to interobserver where Km is the constant relating lumens to
differences in the spectral absorption characteris- watts (683 lumens/watt) and Vk is the spectral
107
■ THE SCIENCE OF COLOR
1.2
1
C3 C2
C1
494 582.5
Chromaticity Coordinate
0.5
C3
C2
460 530 650
C1
0
C3 C2
C2
C1
0.2
400 500 600 700
Wavelength
Figure 3.3 Wright’s chromaticity coordinates plotted as a function of wavelength with WDW normalization.
The solid line represents the average of 10 observers; the dashed lines show the matches of extreme observers.
(From Y. Le Grand, Light, Colour and Vision, translated by R.W.G. Hunt, J.W.T.Walsh, and F.R.W. Hunt, Dover
Publications, New York, 1957; reprinted with permission.)
luminous efficiency of the human eye. The Vk FS FS1 FS2 FS3 l1CS1 l2CS2 l3CS3 (3.12)
had been adopted as a standard by the CIE in
1924 (CIE, 1926), just a few years before For the spectral test lights referred to, an equal
Wright’s experiments. Equation (3.4) can thus energy spectrum, Fk is proportional to Vk.
be rewritten: Further, the luminous unit, lk for a spectral test
light, Ck of luminous flux Fk can be defined:
Ck,i Fk,i/li (3.11)
lk Fk/Ck (l1Ck,1 l2Ck,2 l3Ck,3)/(Ck,1 Ck,2 Ck,3)
where Fk,i is the luminous flux of primary light, (3.13)
(i) specified relative to its luminous unit, li. l1ck,1 l2ck,2 l3ck,3 (3.14)
Wright measured the luminous units for the pri-
maries using heterochromatic flicker photome- If the spectral chromaticity coordinates and
try. The measured values were adjusted to the luminous units, li, of the primaries are
predict the chromaticity coefficients of Standard known, the spectral tristimulus values can be
Illuminant B, consistent with the Vk curve. The derived from the equations above:
relative luminance of the P1, P2, and P3 primaries
was calculated to be in the ratio of 0.65:1:0.044. C k,i(k) ck,iV(k)/(l1ck,1 l2ck,2 l3ck,3) (3.15)
The colorimetric match is a statement of iden-
tity of appearance. At the match, the luminous The data for Figure 3.1 were derived in this
fluxes must also be equivalent. Therefore for any manner from Wright’s tabulation of the chroma-
test light S, of luminous flux FS: ticity coefficients and the values of the relative
108
COLOR MATCHING AND COLOR DISCRIMINATION ■
luminous units given above. In modern termi- where PQ1,1, PQ1,2, and PQ1,3 are the tristimulus
nology, for a set of three primaries, identified as values of primary Q1, PQ2,1, PQ2,2, and PQ2,3 are
P1, P2, P3, the tristimulus values are identified as the tristimulus values of primary Q2, and PQ3,1,
P
Pk,1, Pk,2, and Pk,3, and the chromaticity coordi- Q3,2, and PQ3,3 are the tristimulus values of pri-
nates are pk,1, pk,2, and pk,3. The tristimulus values mary Q3 in the P1, P2, P3 primary set. These
for an arbitrary light, S, are PS,1, PS,2, and PS,3. assignments uniquely determine the transforma-
tion. However, the data will be normalized to the
Linear transformations of color matching energy units of the spectral matches. To obtain
data: Grassmann’s (1853) observation that color the same normalization as the original P1, P2, P3
mixture data show the associative property primary set, a diagonal matrix containing the
allows expression of a set of data in sets of pri- inverse of the energy units, ei is used to multiply
maries other than those used in a particular matrix B.
experiment. This can be noted intuitively since The properties of a homogeneous linear trans-
any test wavelength is a linear combination of a formation (also called an affine transformation)
given primary set (P1, P2, P3). These primaries include: straight lines remain straight after trans-
can themselves be considered as a linear sum of formation; parallel lines remain parallel; and
some other primary set (Q1, Q2, Q3). The equa- plane surfaces remain plane surfaces.
tions are more economically set up in matrix Transformations may also be made directly
notation. between chromaticity coefficients. If matrix B is
A transformation from one set of primaries known, the chromaticity coordinates, q1 and q2,
(P1, P2, P3) to another set (Q1, Q2, Q3) is speci- are given by:
fied by a general homogeneous linear transfor-
mation. The rules for choosing the new q1 b11p1 b12p2 b13p3/[(b11 b12
primaries ensure that the matrix A relating (P1, b13)p1 (b12 b22 b23)p2
P2, P3) to (Q1, Q2, Q3) has an inverse (i.e. the (b31 b32 b33)p3] (3.19)
determinant of A zero). A test color in the
original primary set can be described in terms of q2 b21p1 b22p2 b23p3/[(b11 b12
either primary set. In particular, a given test b13)p1 (b12 b22 b23)p2
color S with tristimulus values PS,i in Primary set (b31 b32 b33)p3] (3.20)
(Q1, Q2, Q3):
The transformation of chromaticity coeffi-
PS,1 a11 a12 a13 QS,1 cients is projective rather than linear. Straight
PS,2 a21 a22 a23 QS,2 (3.16) lines remain straight, but parallel lines need not
PS,3 a31 a32 a33 QS,3 remain parallel.
The subset of physically realizable lights may
has tristimulus values QS,i in Primary set (P1, P2, be thought of as only a small portion of a chro-
P3): maticity space. The possibility exists of defining
new primaries determined by locations in chro-
QS,1 b11 b12 b13 PS,1 maticity space that lie outside the spectrum
QS,2 b21 b22 b23 PS,2 (3.17) locus. In this case, the tristimulus values are not
QS,3 b31 b32 b33 PS,3 known. A transformation matrix, B is deter-
mined, except for an arbitrary scaling factor, by
where matrix B is the inverse of matrix A. In four unique loci in the chromaticity space, the
particular, if the tristimulus values at the new three new primary coordinates and a fourth
primary set (Q1, Q2, Q3) are known in the origi- locus, usually the equal energy spectrum. These
nal set (P1, P2, P3), the matrix A can be written coordinates provide coefficients for three sets of
explicitly: simultaneous equations, one for each new pri-
mary, which are solved to yield the elements of
P
a11 a12 a13 Q1,1 PQ2,1 PQ3,1 matrix B.
a21 a22 a23 P
Q1,2 PQ2,2 PQ3,2 (3.18) Primary transformations have been used for
P
a31 a32 a33 Q1,3 PQ2,3 PQ3,3 widely diverse purposes, including: comparison
109
■ THE SCIENCE OF COLOR
of data sets; derivation of an all-positive (imagi- the G(k). Normalization was to the equal energy
nary) primary system (Judd, 1930); derivation of spectrum. An important consideration was that,
color planes in which equal vector lengths rep- following transform of the chromaticity coordi-
resent equal steps in discriminability (Judd, nates, color matching functions using equation
1935; Galbraith and Marshall, 1985); derivation (3.15) were determined by incorporating V(k),
of a coordinate system in which the primaries the luminous efficiency function of the 1924
represent the cone spectral sensitivities (König Standard observer for photometry. The 1924 CIE
and Dieterici, 1893; Vos and Walraven, 1971; photometric observer was based on an entirely
Smith and Pokorny, 1975; Stockman et al., 1993; different set of observations than the color
Stockman and Sharpe, 2000); and derivation of matching. The accuracy of the 1931 CIE colori-
a coordinate system in which the primaries rep- metric observer is thus dependent both on the
resent physiologically relevant opponent channel accuracy of the Wright and Guild color matching
sensitivities (Schrödinger, 1925; Judd, 1951a). data and on the accuracy of the 1924 CIE spec-
tral luminous efficiency function. Stiles pre-
3.2.3 CIE STANDARD COLORIMETRIC sented color matching data based on absolute
OBSERVERS radiometric calibration. His calculated chro-
maticity coefficients were very close to those of
The industrial importance of trichromacy is in Wright and Guild.
prediction of visual equivalency of a wide array The (X,Y,Z) system is an all-positive system
of spectral power distributions. The colorimetric derived by a linear transformation of the (R,G,B)
data provide a numerical specification at unit system. An all-positive system means that
wavelength steps from which can be calculated the primaries enclose the spectrum locus; the
tristimulus values for any spectral power distri- primaries are not realizable lights and are
bution. The CIE in 1931 defined a standard sometimes called ‘imaginary’ primaries. The
observer for colorimetry, based on 2 color transformation was defined so that the color
matching. The 2 observer is recommended for matching function, Yk was equivalent to V(k),
fields up to 4. A 10 observer was defined in the CIE photopic luminous efficiency function
1964 (CIE, 1964). The characteristics of the for the standard observer. Therefore, the X and
large-field observer are recommended for visual Z primaries have zero luminance. In color space
stimuli whose extent exceeds 4, but should only the line of zero luminance is called the alychne
be used at high photopic illuminances. Revision (Schrödinger, 1920). Given that the luminous
and evaluation of the standard observers units of the (R,G,B) system are in the ratio
remains a current interest of the CIE. 1:4.5907:0.0601, the equation for the alychne
obeys:
The 2 observer: The 1931 CIE standard
observer for colorimetry incorporates both col- r 4.5901g 0.0601b 0 (3.21)
orimetric and photometric behavior. The basic
data were averaged chromaticity coefficients of The X and Y primaries were placed on a line
Wright (1929) and Guild (1931). These were connecting the color matches made above
expressed in Wright’s primaries and normalized 550 nm. This means that Z has no contribution
to Standard Illuminant B. The luminous units of to color matching at long wavelengths. The Y
the primaries were adjusted to be consistent with and Z primaries were placed on a line tangent to
the location of Standard Illuminant B and with the spectrum locus at 504 nm. This choice was to
the luminosity of the 1924 CIE standard minimize the area between the unit triangle and
observer for photometry. the spectrum locus. The intersections of the
Two equivalent statements of the color-match- three lines determine the chromaticity coordi-
ing behavior of the 1931 CIE standard observer nates (r, g) of the new primaries. Normalization
were embodied in the (R,G,B) and (X,Y,Z) was to the equal-energy white. The (X,Y,Z)
systems of units. The (R,G,B) system repre- system is specified by a set of color matching
sented a first step in which spectral primaries functions, Xk, Yk, and Zk shown in Figure 3.4 and a
were retained but all the negative values were in chromaticity diagram, xk, yk shown in Figure 3.5.
110
COLOR MATCHING AND COLOR DISCRIMINATION ■
2.0 1.0
Z(λ)
1.5 0.8
Tristimulus Values
Y(λ) 0.6
1.0
X(λ)
Y
0.4
EES
0.5
0.2
0
400 450 500 550 600 650 700
Wavelength 0
0 0.2 0.4 0.6 0.8 1.0
Figure 3.4 CIE 1931 color matching functions. X
Spectral tristimulus values of constant radiance
stimuli for different wavelengths.The three functions Figure 3.5 CIE 1931 chromaticity diagram. CIE 1931
of wavelength define the color-matching properties of (x,y)-chromaticity diagram. (Plotted from Table 3.1.)
the CIE 1931 standard colorimetric observer. (Plotted
from Table 3.1.)
In 1988, the CIE recommended a supplemen-
tary observer for photometry, termed VM (k).
The (X,Y,Z) system is widely used for color - This observer differs from the Judd revised yJk
specification in industry. Table 3.1 shows color below 410 nm. The further revision to give VM
matching functions and chromaticity coordinates (k) was originally proposed by Vos (1978), who
at 10 nm intervals. also provided tabulations for an amended colori-
metric observer.
The Judd (1951) modified 2 observer (Judd
1951b): After the acceptance of the 1931 CIE The 10 observer: The 1964 CIE large-field
observer, problems were noted in the estimates standard observer for colorimetry was based on
of continuous spectral power distributions. The color matching data from the laboratories of
difficulties were ascribed to underestimation of Stiles and Burch (1959) and Speranskaya (1959,
luminosity at short wavelengths in the 1924 CIE 1961). Since cone-dependent color matching
photometric observer. Judd proposed a revision functions were desired, the experimenters were
of the 1931 CIE colorimetric observer, with a concerned to avoid rod intrusion (changes in
modification of the Yk function below 460 nm color matches caused by rod activity). Stiles and
and a slight modification of the chromaticity Burch (1959) maintained high photopic lumi-
coordinates. The Judd revised colorimetric nance of the matching fields. Multiple primary
observer gave new functions, XJ, k, YJ, k, and ZJ, k sets were used to minimize rod contamination of
in the short wavelength region. The Judd revised the color matches. Finally, Stiles and Burch
colorimetric observer did not replace the CIE 2 (1959) made mathematical corrections to the
observer, which remains widely used in industry. data using a theoretical expectation of the nature
However, the Judd revised colorimetric observer of rod intrusion (see also Wyszecki and Stiles,
is frequently used in color vision theory. Smith, 1982; Shapiro et al., 1994).
Pokorny, and Zaidi (1983) observed that the The CIE transformed the data into an all-
Judd revised colorimetric observer is character- positive system with properties similar to those
ized by less lens and more prereceptoral macular of the 1931 (X,Y,Z) system. The 1964 10
pigment absorption than is the CIE 2 observer. standard observer is specified by a set of color
Table 3.2 shows color matching functions and matching functions, X10(k) Y10(k), and Z10(k) and
chromaticity coordinates at 10 nm intervals for a chromaticity diagram, x10(k), y10(k). The Y10(k)
the Judd (1951) revised observer. represents the relative spectral luminous
111
■ THE SCIENCE OF COLOR
Table 3.1 Color matching functions and chromaticity coordinates for the CIE (1931) Standard colorimetric observer
Wavelength X Y Z x y
380 0.0014 0 0.0065 0.1772 0.0000
390 0.0042 0.0001 0.0201 0.1721 0.0041
400 0.0143 0.0004 0.0679 0.1731 0.0048
410 0.0435 0.0012 0.2074 0.1726 0.0048
420 0.1344 0.0040 0.6456 0.1714 0.0051
430 0.2839 0.0116 1.3856 0.1689 0.0069
440 0.3483 0.0230 1.7471 0.1644 0.0109
450 0.3362 0.0380 1.7721 0.1566 0.0177
460 0.2908 0.0600 1.6692 0.1440 0.0297
470 0.1954 0.0910 1.2876 0.1241 0.0578
480 0.0956 0.1390 0.813 0.0913 0.1327
490 0.032 0.2080 0.4652 0.0454 0.2950
500 0.0049 0.3230 0.272 0.0082 0.5384
510 0.0093 0.5030 0.1582 0.0139 0.7502
520 0.0633 0.7100 0.0782 0.0743 0.8338
530 0.1655 0.8620 0.0422 0.1547 0.8058
540 0.2904 0.9540 0.0203 0.2296 0.7543
550 0.4334 0.9950 0.0087 0.3016 0.6924
560 0.5945 0.9950 0.0039 0.3731 0.6245
570 0.7621 0.9520 0.0021 0.4441 0.5547
580 0.9163 0.8700 0.0017 0.5125 0.4866
590 1.0263 0.7570 0.0011 0.5752 0.4242
600 1.0622 0.6310 0.0008 0.6270 0.3725
610 1.0026 0.5030 0.0003 0.6658 0.3340
620 0.8544 0.3810 0.0002 0.6915 0.3084
630 0.6424 0.2650 0 0.7080 0.2920
640 0.4479 0.1750 0 0.7191 0.2809
650 0.2835 0.1070 0 0.7260 0.2740
660 0.1649 0.0610 0 0.7300 0.2700
670 0.0874 0.0320 0 0.7320 0.2680
680 0.0468 0.0170 0 0.7335 0.2665
690 0.0227 0.0082 0 0.7346 0.2654
700 0.0114 0.0041 0 0.7347 0.2653
710 0.0058 0.0021 0 0.7347 0.2653
720 0.0029 0.0010 0 0.7347 0.2653
730 0.0014 0.0005 0 0.7347 0.2653
740 0.0007 0.0003 0 0.7347 0.2653
750 0.0003 0.0001 0 0.7347 0.2653
760 0.0002 0.0001 0 0.7347 0.2653
770 0.0001 0 0 0.7347 0.2653
780 0 0 0 0.7347 0.2653
efficiency function of the 10 standard observer, triangle completely encloses the experimentally
although it has never been accepted as a standard determined chromaticity diagram (Figure 3.5).
for photometry. Since the data are representative The abscissa (y 0) is the alychne. The spectrum
of matches made by an observer without rod locus forms its characteristic horseshoe shape. A
function, 10 matches made by color normal line connects the coordinates for 380 nm and
observers with other primaries and luminances 700 nm and represents mixtures formed by the
need not in general correspond to the 10 CIE extreme short wavelength and the extreme long
Standard observer. wavelength lights.
Chromaticity coordinates (x, y) may be calcu-
Properties of the chromaticity diagram: In lated for light of any spectral power distribution.
the all-positive (X,Y,Z) system, an isosceles right The tristimulus values of Q are given by:
112
COLOR MATCHING AND COLOR DISCRIMINATION ■
Table 3.2 Color matching functions and chromaticity coordinates for the Judd (1951) Revised colorimetric observer.
Values are tabulated at 10 nm intervals between 380 nm and 780 nm
Wavelength X Y Z x y
380 0.0045 0.0004 0.0224 0.1648 0.0147
390 0.0201 0.0015 0.0925 0.1762 0.0131
400 0.0611 0.0045 0.2799 0.1768 0.0130
410 0.1267 0.0093 0.5835 0.1761 0.0129
420 0.2285 0.0175 1.0622 0.1747 0.0134
430 0.3081 0.0273 1.4526 0.1723 0.0153
440 0.3312 0.0379 1.6064 0.1677 0.0192
450 0.2888 0.0468 1.4717 0.1598 0.0259
460 0.2323 0.0600 1.2880 0.1470 0.0380
470 0.1745 0.0910 1.1133 0.1266 0.0660
480 0.092 0.1390 0.7552 0.0933 0.1409
490 0.0318 0.2080 0.4461 0.0464 0.3033
500 0.0048 0.3230 0.2644 0.0081 0.5454
510 0.0093 0.5030 0.1541 0.0140 0.7548
520 0.0636 0.7100 0.0763 0.0748 0.8354
530 0.1668 0.8620 0.0412 0.1559 0.8056
540 0.2926 0.9540 0.0200 0.2310 0.7532
550 0.4364 0.9950 0.0088 0.3030 0.6909
560 0.597 0.9950 0.0039 0.3741 0.6235
570 0.7642 0.9520 0.0020 0.4448 0.5541
580 0.9159 0.8700 0.0016 0.5124 0.4867
590 1.0225 0.7570 0.0011 0.5742 0.4251
600 1.0544 0.6310 0.0007 0.6253 0.3742
610 0.9922 0.5030 0.0003 0.6635 0.3363
620 0.8432 0.3810 0.0002 0.6887 0.3112
630 0.6327 0.2650 0.0001 0.7047 0.2952
640 0.4404 0.1750 0 0.7156 0.2844
650 0.2787 0.1070 0 0.7226 0.2774
660 0.1619 0.0610 0 0.7263 0.2737
670 0.0858 0.0320 0 0.7284 0.2716
680 0.0459 0.0170 0 0.7297 0.2703
690 0.0222 0.0082 0 0.7303 0.2697
700 0.0114 0.0041 0 0.7347 0.2653
710 0.0058 0.0021 0 0.7347 0.2653
720 0.0029 0.0010 0 0.7347 0.2653
730 0.0014 0.0005 0 0.7347 0.2653
740 0.0007 0.0003 0 0.7347 0.2653
750 0.0003 0.0001 0 0.7347 0.2653
760 0.0002 0.0001 0 0.7347 0.2653
770 0.0001 0 0 0.7347 0.2653
780 0 0 0 0.7347 0.2653
XQ K P(k)
X(k), (3.22) Useful relations to transfer among the tristimu-
YQ K P(k) Y(k), (3.23) lus values and the chromaticity coordinates are:
ZQ K P(k) Z(k), (3.24)
XQ (xQ/yQ)YQ (3.27)
where K is a constant and P(k) is the spectral ZQ (zQ/yQ) YQ ((1 xQ yQ)/yQ)YQ (3.28)
power distribution. The chromaticity coordinates
of Q are given by: In equations 3.22–3.24, K is a normalizing
constant that disappears from the chromaticity
xQ XQ/(XQ YQ ZQ), (3.25) coordinates. In the definition of the 1924 CIE
yQ YQ/(XQ YQ ZQ). (3.26) standard observer for photometry, K is identified
113
■ THE SCIENCE OF COLOR
with km the conversion factor for lumens/watt. occurs in a region of the diagram for which the
The value YQ can thus be expressed in lumens. line extending from A to Q has no intersection
The spectral power distribution Q is completely with the spectrum locus, then the complemen-
specified colorimetrically by its tristimulus values, tary wavelength may be used and should be
XQ, YQ, ZQ, or alternatively by its chromaticity noted by kd or kc. The excitation purity is the
coordinates and luminance, xQ, yQ, YQ. ratio of the distance from A to Q and the distance
from A to kd:
Representation of papers or filters in the
x,y chromaticity diagram: There is an alterna- pe (xQxA)/(xdxA) (yQyA)/(ydyA) (3.32)
tive way to consider K that is useful for specifi-
cation of transmitting filters or reflective In the case of a light whose dominant wave-
surfaces, such as papers. In this case, the colori- length is kd, the intersection with the line
metric properties of the filter or paper under the joining 380 nm and 700 nm is used. Excitation
illuminant are important but the absolute radi- purity is zero when A and Q coincide and
ance level of the source is irrelevant. Suppose unity when Q and kd coincide. Both dominant
there is a pigment sample of spectral reflectance wavelength and excitation purity may be esti-
q(k) or a filter of spectral transmission s(k) mated graphically. Precise computational meth-
viewed under an illumination, H of spectral dis- ods for calculation of dominant wavelength are
tribution H(k). Equations 3.22–3.24 are rewrit- detailed by Wyszecki and Stiles (1982). The
ten to incorporate the spectral properties of the specification of dominant wavelength and exci-
sample, S. For a pigment sample: tation purity is widely used by manufacturers
of colored filters.
XS K q(k)H(k) X(k), (3.29)
k
Primary transformations using the CIE
YS K q(k)H(k) Y(k), (3.30)
k observers: Primary transformations using one
ZS K q(k)H(k) Z(k). (3.31) of the CIE observers are simple because the tris-
k timulus values (X,Y,Z) may be calculated for any
choice of three lights. As an example of a prob-
With K set at 100/H(k)Y(k), YS has the value of
k lem applicable to visual science, consider the use
100[q(k)H(k)Y(k)]/[(k)H(k)Y(k)]. If the object of a color monitor system. The primaries are the
k k
three phosphors. All lights that can be produced
is a perfectly diffusing surface, with q(k) 1.0
on the monitor are combinations of possible light
for all wavelengths, or a perfectly transmitting
outputs of the three phosphors. Calibration of
object, with s(k) 1.0 for all wavelengths, Y has
the relative spectral power distribution will yield
the value 100. Thus, Y may be interpreted as the
the chromaticity coordinates (xi, yi) and calibra-
percentage luminous reflectance or percentage
tion of the maximal luminance will yield Ymax,i
luminous transmittance of the sample. Calcu-
for phosphor (i). The luminance for phosphor (i)
lation of the percentage luminous reflectance or
to light S is described by
transmittance is ratified only for the 1931 CIE
colorimetric standard observer (Wyszecki and
YS,i pS,i Ymax,i (3.33)
Stiles, 1982). Figure 3.6 shows the necessary cal-
culations for a Wratten gelatin (No. 78) and CIE
where pS,i is the proportion of maximal phosphor
Standard Illuminant A1.
output for phosphor (i). Further, the CIE tristim-
A sample whose chromaticity coordinates are
ulus values for light S can be calculated from YS,1,
known may be specified by its dominant wave-
YS,2, and YS,3:
length, kd, and its excitation purity, pe, (Figure
3.6) for a specified CIE Illuminant. The domi-
XS XS,1 XS,2 XS,3
nant wavelength of sample S for Standard
(x1/y1)YS,1 (x2/y2)YS,2 (3.34)
Illuminant A is the wavelength occurring at the
(x3/y3)YS,3
intersection of the spectrum locus and a line
extending from the locus of A through Q. If Q YS YS,1 YS,2 YS,3 (3.35)
114
COLOR MATCHING AND COLOR DISCRIMINATION ■
Wratten No.78 1931 CIE Color
Filter Transmission Standard Illuminant A Maching Functions
1.0 200 2.0
Special Tristimulus Values
Z(λ)
0.8 1.5
Relative Energy
150
Transmission
Y(λ)
0.6 1.0
100 X(λ)
0.4 0.5
50
0.2 0
0 0 –0.5
400 450 500 550 600 650 700 400 450 500 550 600 650 700 400 450 500 550 600 650 700
Wavelength Wavelength Wavelength
Product for Z Product for Y Product for X
25 10 25
20 8 20
Cross Product
Cross Product
Cross Product
15 6 15
10 4 10
5 2 5
0 0 0
400 450 500 550 600 650 700 400 450 500 550 600 650 700 400 450 500 550 600 650 700
Wavelength Wavelength Wavelength
CIE 1931 chromaticity diagram
1.0
0.8
0.6
y
Illuminant A
0.4
+ Kodak # 78
0.2
λ
d
0
0 0.2 0.4 0.6 0.8 1
Figure 3.6 Calculation of tristimulus values and chromaticity coordinates for Wratten filter (No. 78) viewed
under CIE Standard Illuminant A.The upper three panels show the transmission for the filter, the relative
energy output for Standard Illuminant A and the 1931 CIE color matching functions.The middle three panels
show the cross products of the filter transmission, CIE Standard Illuminant A, and the color matching functions.
The lower panel shows the chromaticity diagram with the chromaticity coordinates of CIE standard illuminant
A and the filter-illuminant combination.The line extending from CIE standard illuminant A, through the
chromaticity coordinates of the filter-illuminant combination to the spectrum locus gives the dominant
wavelength of 482 nm.The excitation purity is 0.45 and the luminous transmittance is 8.8%.
ZS ZS,1 ZS,2 ZS,3 where YS,i is given by equation (3.33). It is seen
(z1/y1)YS,1 (z2/y2)YS,2 (3.36) that matrix A contains only the chromaticity
(z3/y3)YS,3 coefficients which are a characteristic of the
phosphors. The inverse matrix gives the lumi-
This can be written in matrix form as:
nances of the phosphors required to produce a
XS x1/y1 x2/y2 x3/y3
YS,1 desired chromaticity and luminance specified in
YS 1 1 1 YS,2 (3.37) CIE coordinates.
ZS z1/y1 z2/y2 z3/y3 YS,3
115
■ THE SCIENCE OF COLOR
3.2.4 EXPERIMENTAL VARIABLES mary was added to the mixture field to allow the
observer to balance the rods in a dark-adapted
The Wright (1929) and Guild (1931) data were state. By repeating the matching procedure in
based on a 2 field viewed with foveal fixation. light- and dark-adapted states, a match is even-
The method was maximum saturation matching tually obtained that holds in both light- and
in which the test light and one primary were dark-adapted states. This ‘tetrachromatic’ match
matched to the remaining primaries in a bipartite does not show a Maxwell spot. Palmer (1981)
field. The effects of procedure, field size, retinal analyzed Trezona’s paradigm and showed that
illuminance, and fixation are important in color her stimulus situation also reduced the differen-
matching experiments and are detailed below. tial effect of macular pigment in the two halves
of the bipartite field.
Maxwell matching: In Maxwell matching, the When the color mixture field is reduced below
spectral light plus a broadband light (such as the 30» of arc, there is a severe loss of discrimination.
EES) are presented in one hemifield and com- With stable fixation, if the field subtends 15» or
pared to the three primaries presented in the 20» and is viewed either foveally or at 20» or 40»
other hemifield (Crawford, 1965). There are from the fovea, the normal trichromatic obser-
subtle differences in the color matching func- ver becomes dichromatic, requiring only two
tions derived from the two methodologies which primaries for full spectrum color matching
have eluded explanation (Zaidi, 1986). Stiles and (Thomson and Wright, 1947). The color matches
Burch (1955) observed that the precision of resemble those of the stationary congenital color
color matches in the short wavelength region of defect, tritanopia (see section 3.5).
the spectrum was improved by superimposing a
mid-spectrum light on both fields. Effect of retinal illuminance: The scalar prop-
erty states that metamers hold for all levels of
Effect of field size: Color matches depend on retinal illumination; however, the range for
the field of view; for example, color matching which the 2 trichromatic metamers hold is
functions based on a 1.5 field (Fridrikh, 1957) limited to about 1–8000 td. Chromatic discrimi-
show systematic differences from the 2 data nation is optimal in a similar range of retinal illu-
(Pokorny et al., 1976). A match for a 2 field will minance. When the retinal illuminance exceeds
not hold for a larger or smaller field. There is a 5000–10 000 td, lights that were metamers are
continuous change in the amounts of the pri- no longer perceived as such (Brindley, 1953;
maries required for the color match as field size Terstiege, 1967; Alpern, 1979; Wyszecki and
is increased (Pokorny and Smith, 1976). Stiles, 1980). For example, in the match of
For certain color stimuli in a 10 colorimetric 589 nm to a mixture of 545 nm and 670 nm
field, a color inhomogeneity may appear at the primaries, a greater proportion of the 670 nm
area of fixation. This appears as an ill-defined primary is required as retinal illuminance
ellipse with major axis horizontal, extending 1 increases above 8000 td. The retinal illuminance
or 2, and is called the Maxwell spot. The spot at which the metamers break down is a level at
follows fixation and is best observed by switch- which photopigment is significantly depleted by
ing rapidly from viewing one half of the bipartite bleaching. While the change in match is a failure
field to the other. In 10 trichromatic matching, of the scalar property, the matches remain
the observer must ignore the Maxwell spot (e.g. trichromatic. The effect and its explanation are
Stiles and Burch, 1959). Alternatively an annu- described further in section 3.2.5.
lar field may be used (e.g. Speranskaya, 1959, With reduction in retinal illuminance, color
1961). The Maxwell spot is usually attributed to matches continue to hold as low as 1 td. One
higher density of macular pigment in the central effect of reduction in retinal illuminance is dis-
1–2 of the fovea. Trezona (1970) advanced an crimination loss, similar to small-field tritanopia
alternate hypothesis that the Maxwell spot (Grigorovici and Aricescu-Savopol, 1958). With
might be a rod contrast color. This conclusion further reduction in luminance, rod intrusion
was based on her studies of 10 color matching becomes evident (Richards and Luria, 1964).
using a tetrachromatic technique. A fourth pri- Whether this is due to rod function in a foveal 2
116
COLOR MATCHING AND COLOR DISCRIMINATION ■
area or to changes in fixation with decreases in ently in the visual system. He proposed three
luminance is unknown. Estévez (1979) sug- physiological mechanisms or fundamentals that
gested that the Wright (1929), Guild (1931), and are differentially sensitive in the visible spec-
Stiles (1955) 2 data have a small rod contribu- trum. In principal, the limitation of trichromacy
tion. Pokorny and Smith (1981) and Pokorny, could occur at any stage in the visual system
Smith, and Went (1981) noted rod contribution (Brindley, 1970). Brindley proposed that foveal
in some color-defective individuals for 50 td trichromacy is photopigment-limited. However,
fields as small as 1. Rod intrusion is more easily with a larger matching field or with parafoveal
seen in the color-defective observer whose cone- viewing, a fourth photoreceptor type, the rod,
dominated color vision is compromised. becomes active. Color matching remains trichro-
matic but does not obey Grassmann’s laws. In
Peripheral color matching: Color matching this case, trichromatic color matching is neurally
may also be performed using the parafoveal or limited (Smith and Pokorny, 1977).
peripheral retina. With the dark-adapted eye and
a scotopic illuminance, the normal observer is The quantal hypothesis: Brindley formalized
monochromatic over most of the spectrum, the biological interpretation of color matching
using the rod mechanism for color matching. At with his proposal that 2 foveal color matching is
mesopic and photopic levels, parafoveal and achieved when the quantal catch rate is equiva-
peripheral color matching is trichromatic. Color lent for each active photopigment (Brindley,
matching data that allow the description of 1970). Since three sets of cone photopigments
peripheral color matches in terms of foveal color are active in the normal 2 fovea, color matching
vision have been described by Moreland and is trichromatic and photopigment-limited. Thus,
Cruz (1959) using an asymmetric matching pro- foveal color matching can be considered a pow-
cedure. For a spectral test field of 40» 80» erful tool in understanding human color vision,
viewed by the periphery and a 40» 80» mixture since the appropriate linear transform will reveal
field viewed by the fovea with WDW normaliza- the effective absorption spectra of the cone
tion (section 3.2.2 above) at the foveal match, visual photopigments. At the same time, color
Moreland and Cruz determined the proportions matching is a limited tool in understanding
of foveal primaries necessary to match various human color vision since it can give us no infor-
peripherally viewed stimuli. The chromaticity mation on the subsequent neural transforma-
coefficients were all inside rather than on the tions performed by the retina and visual cortex.
spectrum locus for Wright’s 2 foveal data, indi- The spectral sensitivity of the fundamentals
cating that a spectral radiation of fixed size must be a linear transform of the color matching
viewed in the peripheral retina is desaturated functions. However, since there are infinitely
in appearance compared with its appearance at many possible linear transformations of color
the fovea. Abramov, Gordon, and Chan (1991) matching functions, external criteria are needed
have emphasized that a peripheral stimulus must to guide and limit the transformation. The classic
be larger than a foveal stimulus in order to appear assumption, made by König was that congenital
colored (see sec 4.2.1 in Chapter 4). Stabell and color-defectives (section 3.5) represented a
Stabell (1976) used the Moreland and Cruz reduction form of normal color vision. An early
paradigm to evaluate rod contribution to color set of fundamentals, which relied on color
matching by taking measurements during the matching data from congenital color-defective
cone-plateau period following intense light adap- observers was published by König and Dieterici
tation and noted that rod participation produces (1886, 1893). Fundamentals that rely on proper-
changes in all three chromaticity coordinates. ties of color-defective vision are called König
fundamentals.
3.2.5 INTERPRETATION OF COLOR
MATCHING Cone fundamentals: König hypothesized that
the dichromatic forms of color defect represent
It was acknowledged by Young (1802) that every reduced forms of normal color vision. Modern
spectral wavelength is not recognized independ- fundamentals are based on the König hypothesis
117
■ THE SCIENCE OF COLOR
(Vos and Walraven, 1971; Smith and Pokorny, nopic copunctal points fall on the inverse diago-
1975; Stockman et al., 1993; Stockman and nal of the spectrum locus in the chromaticity
Sharpe, 2000). We term the fundamentals Sk, diagram, both transformations equate S(k) with
Mk, and Lk to indicate that these fundamentals Z(k). The Smith and Pokorny (1975) transforma-
explicitly represent the energy-based, corneal tion equations, with S(k)/Yj(k) scaled to have a
spectral sensitivities of the short-wavelength height of unity at 400 nm, are:2
(SWS), middle-wavelength (MWS), and long-
wavelength (LWS) sensitive cones respectively. Lk 0.15516 XJ,k 0.54307 YJ,k 0.03287 ZJ,k
Vos and Walraven (1971) suggested that the (3.38)
Judd (1951) revised observer should be used to Mk –0.15516 XJ,k 0.45692 YJ,k 0.03287 ZJ,k
derive fundamentals. They further postulated (3.39)
that the three fundamentals should add to make Sk 0.01608 ZJ,k (3.40)
the Judd (1951) luminous efficiency function,
yJ,k. Smith and Pokorny (1975) suggested that According to the requirement for the transfor-
the luminous efficiency function should be mation:
determined only by L(k) and M(k), giving the
property that the relative height of S(k) is unde- YJ,k L,k Mk (3.41)
termined. Their values for L(k) and M(k) differ
only at short wavelengths from Vos and Table 3.3 shows the Smith and Pokorny funda-
Walraven. Since the protanopic and deutera- mentals as calculated from equations 3.38–3.40.
Table 3.3 Color matching functions and chromaticity coordinates for the Smith and Pokorny fundamentals. Values are
tabulated at 10 nm intervals between 400 nm and 700 nm
Wavelength L M S V l s
400 0.0027 0.0018 0.0045 0.0045 0.6055 1.0002
410 0.0055 0.0038 0.0094 0.0093 0.5948 1.0089
420 0.0100 0.0075 0.0171 0.0175 0.5741 0.9760
430 0.0149 0.0124 0.0234 0.0273 0.5453 0.8556
440 0.0192 0.0187 0.0258 0.0379 0.5059 0.6816
450 0.0219 0.0249 0.0237 0.0468 0.4670 0.5057
460 0.0263 0.0337 0.0207 0.0600 0.4383 0.3452
470 0.0399 0.0511 0.0179 0.0910 0.4385 0.1967
480 0.0649 0.0741 0.0121 0.1390 0.4672 0.0874
490 0.1032 0.1048 0.0072 0.2080 0.4963 0.0345
500 0.1675 0.1555 0.0043 0.3230 0.5185 0.0132
510 0.2695 0.2335 0.0025 0.5030 0.5359 0.0049
520 0.3929 0.3171 0.0012 0.7100 0.5534 0.0017
530 0.4927 0.3693 0.0007 0.8620 0.5715 0.0008
540 0.5628 0.3912 0.0003 0.9540 0.5900 0.0003
550 0.6078 0.3872 0.0001 0.9950 0.6108 0.0001
560 0.6329 0.3621 0.0001 0.9950 0.6360 0.0001
570 0.6355 0.3165 0.0000 0.9520 0.6676 0.0000
580 0.6145 0.2555 0.0000 0.8700 0.7064 0.0000
590 0.5697 0.1873 0.0000 0.7570 0.7526 0.0000
600 0.5063 0.1247 0.0000 0.6310 0.8023 0.0000
610 0.4271 0.0759 0.0000 0.5030 0.8491 0.0000
620 0.3377 0.0433 0.0000 0.3810 0.8865 0.0000
630 0.2421 0.0229 0.0000 0.2650 0.9135 0.0000
640 0.1634 0.0116 0.0000 0.1750 0.9336 0.0000
650 0.1014 0.0056 0.0000 0.1070 0.9472 0.0000
660 0.0582 0.0028 0.0000 0.0610 0.9549 0.0000
670 0.0307 0.0013 0.0000 0.0320 0.9591 0.0000
680 0.0164 0.0006 0.0000 0.0170 0.9620 0.0000
690 0.0079 0.0003 0.0000 0.0082 0.9632 0.0000
700 0.0040 0.0001 0.0000 0.0041 0.9745 0.0000
118
COLOR MATCHING AND COLOR DISCRIMINATION ■
The height of the LWS fundamental at its kmax is MacLeod and Boynton (1979) suggested that a
0.6373; that for the MWS fundamental at its kmax useful and easily used chromaticity chart would
is 0.3924. Figure 3.7 shows the Smith and be a constant luminance plane in which the cone
Pokorny fundamentals, renormalized to their spectral sensitivities formed rectangular axes.
peak and plotted on a logarithmic axis. This is essentially the chromaticity diagram
A recent development is the use of the Stiles formed by the Smith and Pokorny fundamentals
and Burch (1955, 1959) data to derive König with the value of Sk/Yk arbitrarily placed at unity
fundamentals, e.g. (Estévez, 1979; Stockman et at its peak.
al., 1993; Stockman and Sharpe, 2000). The goal
of these attempts was to avoid the criticism that lk Lk/ (Lk Mk) (3.42)
the CIE and Judd standard observers represent mk Mk/(Lk Mk) (3.43)
an amalgam of colorimetric and photometric sk Sk/(Lk Mk) (3.44)
data obtained from different individuals. A new
difficulty is introduced. The Stiles and Burch This cone-based diagram includes the specific
pilot data do not incorporate luminosity, since assumption that sk does not contribute to lumi-
the matching functions were obtained with nance. Figure 3.8 shows the cone-based diagram
direct energy calibrations of the primaries and with sk plotted against lk. In the cone-based dia-
test wavelengths. However, it is now considered gram the horizontal axis represents the
useful for physiological fundamentals to incor- exchange of LWS and MWS cone excitation at
porate luminosity. While the Y of the CIE 10 equiluminance, i.e. an increase in LWS cone
XYZ Standard Observer can be used to represent excitation is offset by a decrease in MWS cone
the relative luminous efficiency function, there excitation but the sum is unity. The vertical
are no measured dichromatic copunctal points axis represents variation in SWS cone excitation
for 10 fields (see section 3.5). at a constant retinal illuminance. The protan
Physiologically based chromaticity diagram:
A physiologically based chromaticity space was 1.0
suggested as early as Maxwell (1860). He postu- 420
lated an isosceles triangle with the cone funda- P D
0.8
mentals at each corner. Although inherently
attractive, the Maxwell triangle is difficult to use. 440
0.6
s/(lm)
0.5 0.4
SWS MWS LWS
0.0 460
Log Relative Sensitivity
0.5 0.2
1.0 480 EES 700
520 600
0
1.5 0 0.2 0.4 0.6 0.8 1
2.0 l/(lm)
2.5 Figure 3.8 MacLeod–Boynton (1979) diagram.
Chromaticity diagram representing relative cone
3.0
excitation.The relative SWS cone excitation s/(lm)
350 400 450 500 550 600 650 700 750 is plotted against l/(lm), the proportion of LWS cone
Wavelength stimulation from equiluminant stimuli.The
isochromatic lines for protanopes P and deuteranopes
Figure 3.7 Relative spectral sensitivity of the cone D converge at l1 (point representing LWS) and m1
visual photopigments as proposed by Smith and (point representing MWS), respectively.Tritanopic
Pokorny (1975).The curves are normalized to their isochromatic lines plot as parallel lines orthogonal to
own maxima. the ordinate.
119
■ THE SCIENCE OF COLOR
copunctal point is at coordinate (0,1); the deutan LQ,k Qk(,Lk/Lmax)(kQ,L/k) (3.49)
copunctal point is at (0,0). Tritan confusions are
represented by a set of parallel, vertical lines. L/Lmax represents the relative spectral sensitivity
Although it is customary to refer to the of the photoreceptor and (kQ,L/k) converts from
MacLeod–Boynton diagram as a cone excitation the energy base of the fundamental to the quantal
diagram, it should be emphasized that the cone base of an absorption spectrum. The term (kQ,L)
excitations are not equivalent to quantal excita- is the wavelength at which the LWS absorption
tion of the cones (see below). spectrum has its peak absorption. The peak of
Rod excitation relative to photopic luminance the absorption spectrum is shifted to shorter
may be incorporated in the cone excitation dia- wavelength than the peak of the fundamental.
gram as a third dimension, by plotting VJk/Vk on When equations 3.48 and 3.49 are combined for
the z axis (Pokorny and Smith, 1986). For chro- each cone type:
maticities produced by three physical primaries,
rod confusion lines for a fixed scotopic lumi- LQ,k Ltd,k/Lmax(107/8) (kQ,L/555) (3.50)
nance will be sets of parallel lines in the equilu- MQ,k QkMk/Mmax) (kQ,M/k)
minance plane (Shapiro et al., 1996). The angles Mts,k/Mmax(107/8) (kQ,M/555) (3.51)
of these lines will be determined by the particular SQ,k QkSk/Smax (kQ,S/k)
choice of primaries. Std,k/Zmax(107/8) (kQ,S/555) (3.52)
Cone trolands: Boynton and Kambe (1980) pro- Thus for a given cone type, the quantal excita-
posed the definition of a new unit, the cone tro- tion at any wavelength is proportional to the
land. Following their suggestion, we can define cone trolands weighted by a constant.
the L, M, and S cone trolands at wavelength, k Second, colorimetric calculations are easily
as: made using the cone troland space (Smith and
Pokorny, 1996). A light, Q specified by the Judd
Lktd (I) Lk/VJk (3.45) revised observer as xJQ, yJQ, YJQ, has a unique
Mktd (I) Mk/VJk (3.46) specification in the cone space given by lQ, zQ,
Sktd (I) Sk 1.6064/VJk (I) Zk/VJk (3.47) YJQ. The tristimulus values, LQ, YJ,Q, ZQ are calcu-
lated in the same way as for XJ,Q, YJ,Q, and ZJ,Q.
where I is the retinal illuminance in trolands, Lk,
Expression in other primary systems follows the
Mk, and Sk are the Smith and Pokorny cone fun-
usual rules (for example, to transform between
damentals, and VJk is the spectral luminous effi-
cone trolands and the phosphors of a color
ciency of the Judd (1951) observer. The SWS
monitor). The equations (3.38–3.40) can be
fundamental is renormalized to be equivalent to
rewritten:
the Zk of the Judd observer (the normalization is
arbitrary in the Smith and Pokorny fundamen-
LQ l1/yJ1 l2/yJ2 l3/yJ3
YJQ,1
tals and was set for convenience in the
YJQ 1 1 1 YJQ,2 (3.53)
MacLeod–Boynton diagram).
SQ s1/yJ1 s2/yJ2 s3/yJ3 YJQ,3
The cone troland has two important proper-
ties. First, cone trolands are proportional to the
where the terms l, yJi, zJi represent the specifica-
quantal excitation rate of the photoreceptors.
tion in chromaticity coordinates for phosphor (i)
Quanta are related to trolands at wavelength k
and YJ, Q,j represents the luminance of light Q for
by the formula:
phosphor (i).
Qk I/ VJk(107/8)(k/555) (3.48)
3.2.6 SOURCES OF INDIVIDUAL
where Qk is the number of trolands, and the DIFFERENCES IN COLOR
other symbols are described above. The quantal MATCHING
excitation rate for a given photoreceptor is
proportional to the number of quanta multiplied The parameter variables that may modify color
by the relative absorption spectrum of the matching data were discussed in section 3.2.4.
photoreceptor. For the LWS cone: These included the field size, the luminance
120
COLOR MATCHING AND COLOR DISCRIMINATION ■
level, the choice of primaries, and the method. unimportant as long as the same are used for
Here we discuss the physiological mechanisms both the data and the synthesized CMFs. Smith,
that may play a role in modifying matches for an Pokorny, and Starr (1976) performed such an
individual or that may play a role in explaining analysis for the Stiles and Burch (1955) 2 color
inter-individual differences in matching. Since matching data and a set of theoretical spectra
color matches are a linear transform of photopig- based in large part on a transformation of the
ment spectral sensitivity, factors that modify the Judd (1951) color matching functions. The
spectral sensitivity can modify color matches. analysis can define the potential extreme limits
These factors include individual variation in of normal variation but cannot specify which
photopigment spectra, variation in optical den- variable or combination of the variables is
sity of the photopigments, photoreceptor optics, responsible for the normal variation. Calculated
and pre-retinal filters. variation in the S, M, and L cone spectral posi-
tions indicated the data variability was domi-
Variation in photopigment spectra: Modern nated by the theoretical L and M cone spectral
molecular biology has defined the protein struc- shifts; S cone shifts had relatively little effect.
ture of the genes coding the photopigment Wavelength shifts in a range from 4 to 2 nm
opsins. To date there is at least one well- in the M cone spectrum and 3 to 7 nm in the
documented polymorphism of the L photopig- L cone spectrum predicted WDW unit coordi-
ment opsin (either the amino acid serine or nates that fell within the extremes of the Stiles
alanine at position 180 on the L- opsin gene; data.
Neitz et al., 1991; Merbs and Nathans, 1992)
which changes the peak wavelength of the Effective optical density of the photopig-
absorption spectrum by a small amount ments: Light must be absorbed in the photopig-
(~3–5 nm). Sanocki, Shevell, and Winderickx ments to be seen. The concentration of the
(1994), using a technique which eliminates in pigment and the length of the light path affect an
large part inter-observer differences in prerecep- absorption spectrum. As light traverses a greater
toral filtering and photopigment optical density, concentration of pigment, more light is absorbed
show reliable small differences in color matching and the extinction spectrum broadens. The rule
in the red–green spectral region for observers governing absorption follows Beer’s law. In the
having the two different alleles. The same case of a visual photopigment, expressed in
allele pattern also is present for the M-cone decadic base:
photopigment but at a low frequency and it is
probable that there are other polymorphisms A (1–10–kcl) (3.54)
capable of modifying extinction spectra by
small spectral shifts. There is now accumulat- where A is the fraction absorbed, k is the decadic
ing evidence that individuals with normal color extinction coefficient, c refers to the concentra-
vision exhibit small variation in the absorption tion and l to the pathlength. The extinction coef-
spectra of the photopigments in their L and M ficient, k, is a wavelength-dependent function
cones. that is characteristic of the photopigment. For
An estimate of the extreme values that pho- visual photopigments, concentration and path-
topigment variation and optical density might length can be grouped as a single factor, the
assume in the normal population may be effective optical density of the photopigment.
obtained by theoretically manipulating these A graphical sketch to demonstrate how the
variables and establishing the extreme values absorption spectrum broadens as the effective
that fall within the intra-observer variability of optical density increases is shown in Figure 3.9.
color matching data. If WDW normalization The insert in the top panel schematically illus-
(Wright, 1946; Wyszecki and Stiles, 1982) is trates a path through a beaker of the rod pho-
used for both data and the synthesized CMFs, topigment rhodopsin. The curves show the
the effects of individual differences in prerecep- calculated effective fractional absorbance at each
toral filtering are eliminated. The choices of pri- successive layer of pigment. Light reaching each
maries and normalizing wavelengths are successive layer is the product of the incoming
121
■ THE SCIENCE OF COLOR
the absorbances of all layers. The middle panel
12 I
shows the fractional absorption spectrum, calcu-
1
lated for different optical densities of pigment.
Light Absorbed (percent)
10 1 .0
2 .10 The lower panel shows the relative fractional
2 3 .25 absorption.
8
4 .45 The analysis proposed by Smith, Pokorny and
3
6 Starr can also be applied to variation in optical
5 .70
density. Calculated variation in photopigment
4
4 6 1.00 optical densities could also predict the variability
5 of the Stiles and Burch data. In this analysis a
2 6 higher optical density was required for the L
cone photopigment, a result found for all pub-
0
420 460 500 540 580
lished comparisons of L and M cone optical den-
sity (Miller, 1972; Smith and Pokorny, 1973;
∞ Bowmaker et al., 1978; Wyszecki and Stiles,
100
1.0 1980; Burns and Elsner, 1993). An optical
density range of 0.15 to 0.45 for the M cone
Light Absorbed (percent)
0.5 spectrum and 0.25 to 0.55 for the L cone spec-
trum predicted WDW unit coordinates that fell
50 within the extremes of the Stiles (1955) data.
0.2 The estimated standard deviation was 0.05.
Cone photoreceptors vary in length, with reti-
nal position being longest in the center of the
0.0 fovea (Polyak, 1957). Field size is an important
0 variable in matching which may be attributed
∞ partially to changes in the effective optical den-
100
sity of photopigments. Pokorny, Smith and Starr
(1976) extended their analysis to the difference
Percent of Maximum
of the Stiles (1955) 2 and the Stiles and Burch
(1959) 10 data. The differences between the
50
mean 2 and 10 data could be explained by a
1.0 reduction in optical density of 0.10–0.15.
0.5 Pokorny, Smith and Starr also calculated the the-
0.2
0.0 oretically acceptable range of optical densities of
the L and M cone photopigments for data col-
0 lected with a small field at several eccentricities.
Thomson and Wright (1947) obtained small
400 500 600
field tritanopic data from W.D. Wright’s eye
Wavelength (nm)
with a 15» field fixated directly or placed at
Figure 3.9 Effect of optical density on spectral eccentricities of 20» and 40» from fixation. There
sensitivity. (The top panel is from Goldstein and have been a number of experimental verifica-
Williams, 1966; the bottom two panels are from tions of the change in color matches with field
Dartnall, 1957.) size (Horner and Purslow, 1947; Pokorny and
Smith, 1976; Burns and Elsner, 1985; Eisner et
al., 1987; Swanson and Fish, 1996). These
light and the transmission of the layers that studies are all consistent with the interpretation
precede them. At the initial layer (1), the absorp- that the effective optical densities of the photo-
tion resembles that of the extinction spectrum. pigments decrease as field sizes increase. Figure
At the final layer (6), most of the absorption is 3.10 shows the estimated L cone optical density
taking place in the tails of the extinction spec- as a function of field size from a number of
trum. The absorptance spectrum is the sum of studies.
122
COLOR MATCHING AND COLOR DISCRIMINATION ■
(A) 2.0
1.0 W0
1.5
Effective Optical Density
Optical Density
0.8
1.0
0.6
0.4 0.5
0.2
0
400 450 500 550 600 650 700
0 Wavelength
0 2 4 6 8 10
Stimulus Diameter (deg) (B)
0.5
Figure 3.10 Optical density of the L cones as a 0.4
function of field size. (From Pokorny et al., 1976.) Optical Density
0.3
Pre-retinal filters: The other major sources of 0.2
variation in color matching are pre-retinal filters,
namely the lens and macular pigment density. 0.1
Figure 3.11 shows the density spectra for an
average lens of a 32-year-old observer and the 0
average macular pigment for a 2 field. These both
show substantial variation in the population. 400 450 500 550
There is significant inter-individual variation Wavelength
in lens transmission and also a substantial
Figure 3.11 (A) The optical density of the lens (van
change with age. Van Norren and Vos (1974) Norren and Vos, 1974) and (B) the macular pigment
calculated from Crawford’s (1949) scotopic spec- (Wyszecki and Stiles, 1982).
tral sensitivity data that individual variation in
ocular absorption for a group of 50 young
observers (17–30 year) is about 25% of the eliminate those with incipient cataracts (Weale,
average absorption at short wavelengths. Adult 1988; Moreland et al., 1991).
lens transmission has been characterized as hav- The macular region of the human retina con-
ing two components, TL1 and TL2 with TL1 varying tains a non-photolabile pigment that selectively
with age (Tan, 1971; Pokorny et al., 1987). An filters light arriving at the receptors. Entopic
algorithm that allows calculation of an average viewing of Maxwell’s spot (section 3.2.4) is inter-
lens transmission between the ages of 20 and 60 preted as a visualization of the spatial distribution
years is: of macular pigment (Palmer, 1978). For many
observers Maxwell’s spot is seen to be darkest in
TL TL1 [1 0.02(A 32)] TL2 (3.55) the center of the fovea but for others there is con-
siderable spatial structure (Miles, 1954).
where TL, TL1, and TL2 represent the spectral lens Studies of the optical density of the macular
densities tabulated in Table 3.4 and A is the pigment have used widely different methodolo-
chronological age. Beyond the age of 60, there gies. Since macular pigment varies with field
is an acceleration in decrease in lens (Pokorny position (see below), it is necessary when evalu-
et al., 1987). This acceleration is characteristic of ating literature reports to take into account the
a population average but was not observed in field size employed in the various studies. Perhaps
studies where observers were screened so as to the most striking feature of the literature reports
123
■ THE SCIENCE OF COLOR
Table 3.4 Tabulation of the Optical Density of the Total Lens Transmission Function TL for an Average 32-year-old
Observer and Separation of TL into Components:TL1 Represents Portion Affected by Aging after age 20, and TL2
Represents Portion Stable after age 20.* (After Pokorny, Smith and Lutze, 1987)
Optical Density Optical Density
Wavelength Wavelength
nm TL TL1 TL2 nm TL TL1 TL2
400 1.933 0.600 1.333 530 0.120 0.120 –
410 1.280 0.510 0.770 540 0.107 0.107 –
420 0.787 0.433 0.354 550 0.093 0.093 –
430 0.493 0.377 0.116 560 0.080 0.080 –
440 0.360 0.327 0.033 570 0.067 0.067 –
450 0.300 0.295 0.005 580 0.053 0.053 –
460 0.267 0.267 – 590 0.040 0.040 –
470 0.233 0.233 – 600 0.033 0.033 –
480 0.207 0.207 – 610 0.027 0.027 –
490 0.187 0.187 – 620 0.020 0.020 –
500 0.167 0.167 – 630 0.013 0.013 –
510 0.147 0.147 – 640 0.007 0.007 –
520 0.133 0.133 – 650 0.000 0.000 –
* The optical density of the lens of an average observer between the ages of 20 and 60 years old may be estimated by
TL TL1 [1 0.02(A-32)] TL2.
For an average observer over the age of 60
TL TL1 [1.56 0.0667(A-60)] TL2,
where A is the observer’s age.
TL is the van Norren and Vos (1974) tabulation of lens density scaled by 1.333 to represent a 32-year-old observer (the
average age of the Stiles and Burch observers) with a small pupil (<3mm). To estimate the lens density function for a
completely open pupil (>7mm), multiply the tabulated values by 0.86.
For the Wyszecki and Stiles (1982) lens density function, the following values may be substituted:
Wavelength Optical Density
nm TL TL1 TL2
400 1.600 0.600 1.000
410 1.093 0.510 0.583
420 0.733 0.433 0.300
is the marked individual variation in macular Stabell, 1980; Viénot, 1983; Hammond et al.,
pigment; for a 2 field size, individuals may vary 1997). These data may be characterized with an
from 0.0 to 1.0 optical density at 460 nm, the exponential decrease in macular pigment optical
wavelength of peak absorption (Vos, 1972; Pease density with retinal eccentricity (Moreland and
et al., 1987; Werner et al., 1987). Bhatt, 1984; Hammond et al., 1997).
Psychophysically measured density of the
macular pigment decreases with increase in field
size. Estimates from literature reviews of the 3.3 CHROMATIC DETECTION
average macular pigment optical density for a
2 field vary from 0.35 (Vos, 1972) to 0.50 Detection refers to the ability to recognize a
(Wyszecki and Stiles, 1982). Stiles (1955) change in light level. The threshold for detection
reported that the 10 colorimetric data are char- can be measured in the fully dark-adapted eye
acterized by macular pigment densities of about (absolute threshold) or on a background (incre-
0.25 times the 2 optical density. Four studies ment threshold). The threshold represents some
have attempted to characterize the retinal distri- criterion change presumably correlated with a
bution of macular pigment by comparison of change in a neural response from its steady-state
data taken for a small field at a number of reti- adapting level. It is possible to specify absolute
nal eccentricities (Ruddock, 1963; Stabell and threshold in terms of the number of quanta
124
COLOR MATCHING AND COLOR DISCRIMINATION ■
needed for detection. The increment threshold is will be mediated by cone mechanisms. The data
usually expressed either as the increment or may be expressed in several different metrics.
change in quanta, Q, from the background The detection threshold can be plotted as the
level or as the contrast ratio of the increment to log R versus log R and the function is called
the background, Q/Q. Increment thresholds a threshold-versus-radiance (TVR) function.
may equally well be expressed in quantal, Alternately, the data may be expressed in units
radiance or luminance units. of retinal illuminance and the function is called
Sensitivity is inversely related to detection; it is a threshold-versus-illuminance (TVI) function.
the reciprocal of the quanta at the detection Data may also be specified in quanta/deg2/sec.
threshold. The human visual system maintains The TVR function shows a characteristic shape.
an approximately even state of contrast sensitiv- At low radiances, detection threshold is un-
ity over a billion-fold range of light levels. changed from the absolute threshold on a zero
Detection thresholds can be measured over this background. As the background is increased,
entire range of illumination. In the mid- threshold starts to increase and reaches a limiting
periphery of the dark-adapted eye, where the R/R slope of unity. At this limiting slope, sensi-
visual system is most sensitive, the absolute tivity is said to be in the Weber region; threshold
detection threshold requires only a few quanta. is a constant percentage of the background, or a
The visual system continues to maintain sensi- constant contrast.
tivity even at steady-state background levels of
1011.5 quanta (about 20 000 td at 555 nm). At 3.3.2 EXPLANATORY CONCEPTS IN
this quantal excitation level, there is a 50% DETECTION
depletion of the MWS and LWS cone photo-
pigments. At such levels, an increment of many In the psychophysical literature, there have been
millions of quanta is needed to detect a change three important explanatory concepts in describ-
from the steady excitation level. ing the TVR function. These are adaptation, sat-
Part of the sensitivity range of the human uration, and noise.
visual system is obtained by the duplex nature of
the retina. There are two sub-systems each with 3.3.2.1 Adaptation
an operating range over a million-fold, that Adaptation refers to mechanisms that expand
show overlapping ranges of sensitivity. One the dynamic range of response of the system as a
receptor system, the rods, constitutes a high sen- whole. Any given retinal element has a limited
sitivity, low visual acuity, color-blind system. The response range of perhaps 400-fold. Thus, with-
other system, the cones, represents a high out adaptation, as light level increases the range
threshold, high visual acuity, color system. There available for greater stimulation decreases.
are a variety of sources which treat the topics of Following adaptation, a neural element may
detection and sensitivity more fully (Cornsweet, maintain this 400-fold response range for a wide
1970; Barlow, 1972; Hood and Finkelstein, 1986). range of background light levels. The literature
Here we briefly review some explanatory con- distinguishes two major sub-types of adaptation.
cepts in detection and review chromatic detec- Subtractive adaptation resets the steady-state
tion data aimed at isolating cone mechanisms signal without affecting the response to a stimulus
and/or postreceptoral channels. pulse. Multiplicative adaptation scales both the
responses to the steady-state and to a stimulus
3.3.1 THRESHOLD-VERSUS-RADIANCE pulse. Multiplicative adaptational mechanisms
(TVR) FUNCTIONS are sometimes called gain controls.
Multiplicative adaptation has many sources.
Detection thresholds can be measured as a small, The reflex response of the closing of the pupil in
brief increment, R on a larger steady back- bright illumination is a source of multiplicative
ground of radiance R. At each background level, adaptation in the natural environment. Pupil
the observer first adapts to the background. If constriction reduces the amount of retinal illu-
the test and background are foveally fixated and mination from both a steady-state background
the spectral distribution is uniform, detection and from a light increment. Pupil constriction
125
■ THE SCIENCE OF COLOR
allows a 16-fold increase in the total operating It is common to lump sources of multiplicative
range of the visual system, under conditions of adaptation in a single term. At a computational
natural viewing and thus plays a relatively level, multiplicative adaptation is a scalar less
minor role in adaptation in the natural environ- than one that modifies the adapted response, RA
ment. In threshold experiments, the effective to both the background, QA and the increment,
pupil is often held constant by use of an artificial Q. The response to the background is given by:
pupil (Troland, 1915; Pokorny and Smith, 1997)
allowing the study of neural adaptational mech- RA QA/(1 kQA) (3.57)
anisms. An artificial pupil is an aperture smaller
that the natural pupil which the observer looks where k is an adaptation constant. Provided that
through. An alternate technique sometimes used the threshold increment does not perturb the
in Maxwellian view systems places a limiting adaptation level, the increment threshold is
aperture in a plane conjugate to the natural described by the equation:
pupil.
Photopigment depletion is another source of Q (Qth) (1 kQA) (3.58)
multiplicative adaptation. Breakdown and re-
generation of photopigment are reciprocal pro- where (Qth) is the absolute threshold, deter-
cesses. At low illuminations, regeneration works mined by criterion, noise, and/or quantal
to maintain a full complement of photopigment. requirements. The adaptation constant, k, is the
At high illuminations, the visual photopigment is reciprocal of the quanta at which threshold is
depleted (sometimes termed bleaching). In steady raised two-fold. On a double logarithmic scale,
illumination, there is no net effect. For a first order the limiting slope is unity, as demanded in the
kinetic process (Rushton, 1972) the amount of Weber region. This equation describes the data
photopigment depleted (1p) is given by: of TVR functions obtained for the cone mecha-
nisms on large steady backgrounds. The function
(1p) I/(I Io) (3.56) is sketched in Figure 3.12.
Subtractive adaptation is considered to
where p is the amount of photopigment avail- remove part of the adapting background
able, I is the retinal illuminance and Io is a con- (Geisler, 1981; Adelson, 1982; Hayhoe et al.,
stant, the illumination at which 50% of the 1987). When subtractive adaptation is added to
photopigment is depleted. For cone photopig-
ments at their kmax, the value of Io is 1011.5 quanta 10
(about 20,000 td at 555 nm). At levels above
20,000 td, there is sufficient depletion that only
a lower percent of the incident background 8
quanta are absorbed. If the system is responsive Saturation
at low bleaching levels, it will remain responsive
Log Delta R
6
at higher levels since further increases in inci-
dent background quanta will be offset by the
reduced probability of absorption. This mecha- 4 Weber
nism of adaptation usually plays only a small
role in the natural environment, however, snow
2
fields and sunlit beaches may provide sufficient Noise
illumination to allow substantial bleaching.
The most important sources of multiplicative 0
adaptation occur neurally. Recordings from pri- 2 1 0 1 2 3 4 5
mate horizontal cells, second-order neurons in Log Radiance
the retina show adaptation over a wide range of
Figure 3.12 The increment threshold functions
light levels (Smith et al., 2001) though the extent predicted by three types of detection mechanism, a
of adaptation is not as great as found later in the saturation function, adaptation following Weber’s law,
retina (Lee et al., 1990), or psychophysically. and a noise mechanism.
126
COLOR MATCHING AND COLOR DISCRIMINATION ■
a process already controlled by multiplicative Saturation can be found in cone-mediated
adaptation, the effect is to shift the TVR function TVR functions under special conditions of meas-
on the horizontal axis. urement. Protocols that reveal saturation
include pulsed backgrounds (King-Smith and
3.3.2.2 Saturation Webb, 1974; Shevell, 1977; Stockman et al.,
Any given neural element has a limited response 1993) and the probe-flash technique (Hayhoe
range of perhaps 400-fold. The response of a et al., 1987; Hood, 1998). A large steady back-
non-adapting retinal element initially increases ground controls the overall state of adaptation.
with the incident quanta, but with diminishing An adapting pulse of a second or so, called the
slope. At a high level above threshold, the flash is superimposed on the background, and a
response is constant and is said to have satu- brief test pulse, the probe is superimposed on the
rated. In vision, the saturating response, R is adapting flash. The probe may be presented at
usually described by a Naka–Rushton equation: any time interval during the flash presentation.
If the probe is presented in the middle of the
R Rmax (Q)/(Q Qsat) (3.59) flash duration, threshold follows a Weber func-
tion, indicative of neural adaptation to the
where Rmax is the maximal response, Q is the flash. If the probe is presented at flash onset,
incident quanta and Qsat is the semi-saturation, threshold is elevated above the Weber function
or number of quanta at which the response is demonstrating a slope of two before saturation,
half the maximum value. It is clear from equa- indicative of a saturating process. The probe-
tion (3.60) that sensitivity is greatest at low light flash technique has been used to delineate a
levels, and that at high light levels where the different time course for subtractive and
response has saturated, a detectable increment multiplicative adaptation (Geisler, 1981; Hayhoe
in response will not be measurable. The equation et al., 1987).
for detection for a saturating system obeying the
Naka–Rushton equation is: Mechanisms of neural adaptation: The neu-
ral multiplication and neural subtractive adapta-
R (d/Rmax) (Q Qsat)2/(Qsat tion inferred from the data described above has
(d/Rmax) (Q Qsat)) (3.60) computational importance. Another approach is
to ask what kind of neural mechanism might
where d is the criterion, Rmax is the maximal
exist. Neural mechanisms may be feed-forward
response, Q is the incident quanta and Qsat is the
or feed-backward. Historically, feed-forward is
semi-saturation, or number of quanta at which
considered unsatisfying since it must be subject
the response is half the maximum value.
to saturation.
Normally the term (d/Rmax) is very small. At
In subtractive feedback, a portion, k of the
absolute threshold, where Q is zero:
signal is subtracted from the signal. In the steady
R ~ (d/Rmax) (Qsat) (3.61) state subtractive feedback is described as
At higher quantal excitation levels where R Q (kR) (3.63)
Q Qsat: R Q/(1k) (3.64)
R ~ (d/Rmax) (Q2) (3.62) Subtractive feedback acts to shift the start of a
saturating mechanism to a higher radiance level.
This equation then describes a function that has Models of subtractive feedback may postulate
a slope of two, steeper than the limiting slope of that the feedback is developed slowly in time
the Weber function. For some values of d, Rmax compared with the signal. Slow subtractive feed-
and Qsat, the denominator of equation 3.60 back shows pronounced overshoots following a
becomes negative, implying that negative light is rapid increase in radiance level.
required at threshold. It is at this point that the In divisive feedback the signal is divided by
system is said to have saturated. Equation 3.60 is a portion of the signal. In the steady state
sketched in Figure 3.12. response is given by:
127
■ THE SCIENCE OF COLOR
R Q/(1kR) (3.65) clusion from this study was that a single quan-
R 1⁄2k [(14kQ)/4k2]0.5 (3.66) tum is sufficient to excite a photoreceptor.
However, the study also suggested that on aver-
In divisive feedback, the response increases as age five to eight absorptions were required for
the square root of the signal; i.e. only part of the detection on 55% of the trials. Thus detection is
signal is removed by the adapting mechanism. limited not by absorption but by some neural cri-
Subtractive feedback acts to shift a saturating terion. Barlow (1956) introduced the idea that
mechanism by the square root of the steady-state the photoreceptor responses might themselves
radiance. be noisy in the absence of stimulation. This idea
is sometimes called ‘equivalent noise,’ EN.
Combined neural mechanisms: The subtrac- Equation (3.60) can be rewritten:
tive and divisive feedbacks described above do
not produce Weber behavior. There is interest in R (EN Q)0.5 (3.68)
the literature as to how Weber behavior may
be produced. One possibility, as proposed by In the absence of a real background, absolute
Koenderink, van de Grind and Bouman (1970) threshold will be limited by the equivalent noise.
was to weight a subtractive feedback pathway Once the background noise exceeds the neural
with a feed-forward signal that represented the noise, detection will be limited by the back-
average radiance over time. ground noise. The limiting slope of a noise-
limited TVR function is 0.5 (see Figure 3.12).
3.3.2.3 Noise The noise concept has been most important at
An important concept in detection is that the sig- absolute threshold. As the steady-state light level
nal may be noisy. There are two consequences of increases, multiplicative adaptation processes
the noise concept in visual thresholds. First, take over and detection is in the Weber region.
noise affects the slope of the psychometric func- Provided the multiplicative adaptation process
tion and, second, a noise-limited process affects follows the source of the noise, it will reduce the
the slope of the TVR function. The applicability effect of both the average background and its
of the noise concept at absolute threshold fol- variance. Thus, the level of stimulus noise or
lowed the recognition that the statistics of light early neural noise will approach a constant value
emission in a brief pulse follow a Poisson once gain mechanisms are active. It is also rec-
process. The introduction of the noise concept in ognized that threshold may be limited by late
visual processing occurred at a time when engi- neural noise that follows gain mechanisms.
neers were developing methods to assess signal There are data that suggest that the intrinsic
transmission in electronic transmission, i.e. sig- noise of spike generation at the retinal ganglion
nal detection theory. A detection event then cell is constant at different superthreshold illu-
requires that the response to a signal on a noisy minations (Troy and Robson, 1992; Troy and
background be some criterion amount larger Lee, 1994).
than the noise level alone. If the noise is Poisson- There are conditions for which cone-mediated
limited, its variance is given by the mean. detection obeys equation (3.68) (Barlow, 1957;
Suppose threshold is defined at 55% correct Bouman and Koenderink, 1972). The conditions
detection, i.e., detection occurs at one standard to obtain noise-limited detection in cones
deviation above the noise. The increment involve the use of very small test and adaptation
threshold is defined as: fields. These conditions probably restrict the role
of neural adaptation.
R (Q0.5) (3.67)
3.3.3 DETECTION ON SPECTRAL
In the classic paper of Hecht, Shlaer, and Pirenne BACKGROUNDS
(1942), the slope of the psychometric function at
rod absolute threshold was related to the uncer- The TVR function for foveally fixated achromatic
tainty in the number of quanta delivered per targets measured with large steady backgrounds
light pulse. An important and undisputed con- and small test stimuli follows the form of
128
COLOR MATCHING AND COLOR DISCRIMINATION ■
equation (3.58). However, when chromatic cept and the term k for adaptation determines
backgrounds and tests are used, multiple the transition to the Weber region.
branches or partial components of multiple TVR
functions can be measured depending on the The vertical displacement law (test wave-
choice of test and background wavelengths. length variation): Suppose test and back-
Further, the spectral sensitivity of the detection ground are chosen near the peak absorption of
mechanisms varies, depending on the back- the photopigment. At absolute threshold, the
ground wavelength. These results were attrib- mechanism is at its most sensitive, since the cho-
uted to multiple underlying cone mechanisms sen wavelength allows the highest probability of
(i.e. a photoreceptor and its adaptation mecha- absorption. If the test wavelength is now
nism). An early goal of these studies was to sep- changed, the absolute threshold will be higher,
arate and measure the spectral sensitivity of the since absolute threshold is determined partially
cone mechanisms with the view that they would by the reciprocal absorptivity of the underlying
reveal the underlying visual photopigments. photopigment (Chapter 2). The transition to
One important pre-theoretical concept was the Weber region, however, is determined by the
that a spectral background might exert a differ- relative absorptivity of the photopigment at the
ential effect on the three classes of cones. The adapting background that remains unchanged.
differential spectral sensitivity of the cone pho- The TVR functions for different test wavelengths
topigments might be exploited to allow isolation will be displaced vertically, but will make their
of each cone mechanism by an appropriate transition into the Weber region at the same
choice of background wavelength e.g. Wald background radiance. The vertical displacement
(1964). This idea in turn originated in the law states that variation in test wavelength can
photochemical studies of extracted pigments only affect the vertical position of the TVR
(Dartnall, 1957). It was recognized that the con- function for a single mechanism.
stituents of extracted photopigment mixtures
could be studied by partial bleaching, i.e. by The horizontal displacement law (back-
bleaching with a spectral wavelength that ground wavelength variation): In compari-
exploited the differential transmissivity of the son, consider variation in the background
pigment mixture. In human psychophysics, it wavelength, leaving the test wavelength fixed
became clear that this exploitation could be near the peak absorption of the photopigment.
achieved at levels below photopigment deple- The value of absolute threshold is not affected by
tion. These data suggested that the neural adap- the choice of background wavelength. However,
tation mechanisms of the cone mechanisms the effectiveness of the background, QA is deter-
were equivalently independent. The major mined by the absorption probability of the pho-
development of the relation between cone- topigment. The transition to the Weber region
mediated TVR functions and the underlying occurs at higher radiances for background wave-
cone mechanisms was performed by Stiles lengths other than at the peak absorptivity of the
(1959, 1978). photopigment. The TVR functions will be dis-
placed horizontally on the radiance axis but will
3.3.3.1 Displacement laws share the same vertical position since they are
When TVR functions are measured using mono- pinned at absolute threshold. The horizontal
chromatic lights for test and background stimu- displacement law states that variation in back-
lus, there are two important laws which govern ground wavelength can only affect the horizon-
the placement of the TVR function on the hori- tal position of the TVR function for a single
zontal or vertical axis. Consider a single mecha- mechanism.
nism consisting of a photopigment with neural It is clear that the spectral sensitivity of a sin-
elements that show multiplicative adaptation, gle mechanism can be assessed in two different
giving a classical TVR function. From equation ways. Spectral sensitivity can be assessed by
(3.58), we see that two independent factors using a background of fixed wavelength and a
govern the TVR function. The term Qth for test stimulus of varying wavelength. The back-
absolute threshold determines the vertical inter- ground radiance is fixed and test radiance is
129
■ THE SCIENCE OF COLOR
varied to achieve increment threshold. This is wavelength. He termed these p mechanisms
called a test sensitivity function. Alternatively, (Figure 3.13). These were subsequently repli-
spectral sensitivity can be assessed by using a test cated and subjected to tests of independence
stimulus of fixed wavelength and a background (Pugh and Kirk, 1986).
stimulus of varying wavelength. The radiance of p 1,2,3: The first three Stiles p mechanisms are
the background is varied to achieve a 10-fold rise associated with the SWS cone mechanism. The p
in test wavelength threshold above absolute 2 mechanism is the most sensitive mechanism,
threshold. This spectral sensitivity function is obtained at absolute threshold. The p 2 mecha-
called a field sensitivity function. The two meth- nism occurs variably among observers, and thus
ods each have advantages and disadvantages. has not been the subject of extensive analysis.
The test sensitivity method is more straight- The p 1 mechanism is obtained at low to moder-
forward, and requires minimal data collection ate adapting levels. The p 3 mechanism is
(one threshold per test wavelength). Test sensi- obtained when adaptation fields are close to
tivity has seen the greater use (Wald, 1964; bleaching levels for the LWS and MWS cones.
Eisner and MacLeod, 1981; Yeh et al., 1989). Its One difficulty in obtaining the p 1 and the p 3
disadvantage is that the human cone mecha- mechanisms is that even with a 435 nm test
nisms are highly overlapping, while the success flash near the peak of the SWS cone spectral
of the technique depends on a large differential sensitivity, an auxiliary long wavelength condi-
sensitivity between constituent mechanisms. It is tioning field is needed to suppress the other cone
virtually impossible to choose a background types. The finding of multiple mechanisms for
wavelength that will yield test spectral sensitiv- SWS cone function was disappointing in that it
ity for the entire spectrum for a given cone revealed a failure to isolate a unitary mechanism
mechanism. The field sensitivity method is more that could be associated with a presumed SWS
data-intensive; a TVR function must be meas- cone photopigment. Further, the long wave-
ured at each background wavelength. The over- length sensitivity slopes of the p 1, the p 2, and
lap of the human cone spectral sensitivities may the p 3 mechanisms were not consistent with
result in overlap of the TVR functions. Thus, a typical absorption functions for photopigments.
value for the spectral sensitivity at 10 absolute These findings led Stiles to realize that his
threshold may in some cases depend on extrap- techniques did not reveal a unitary adapting
olation of a fitted TVR function rather than a
direct measurement.
Wavelength
400 500 600 700
The p mechanisms: If the cones have inde- 1
pendent adaptation pathways, then the displace-
4
ment laws will hold for each independent 0
1 5
mechanism. Additionally, provided the adapta-
Log Relative Sensitivity
tion mechanism is such as to generate a TVR 1
function showing Weber’s law, then the spectral
2
sensitivity of the underlying mechanism may be
inferred even if only a small segment of the TVR
3
function is measurable. It is also possible to test
the independence of the adapting mechanisms, 4 3
by superimposing backgrounds that are mixtures
of component wavelengths. Stiles developed the 5
rationale of the displacement laws and the field
sensitivity technique. Stiles used a consistent 6
experimental paradigm. The background was 10 25 000 20 000 14 000
and the test was a 1 by 1 square pulsed for Wavenumber
100–200 msec. Stiles obtained a number of func- Figure 3.13 Stiles field mechanisms. Log relative
tions of differing spectral sensitivity, associated quantal field sensitivities of the p mechanisms plotted
with different combinations of test and field as a function of wavenumber.
130
COLOR MATCHING AND COLOR DISCRIMINATION ■
mechanism. Subsequent studies provided fur- modern models of adaptational mechanisms, the
ther indication that the p 1 to p 3 are not consis- gain controls are carried forward to subsequent
tent with a unitary adapting mechanisms (Pugh, sites. An early stage of multiplicative adaptation
1976; Mollon and Polden, 1977; Pugh and serves to stabilize the later opponent sites and
Mollon, 1979). In particular, there were failures thus the first and second sites cannot be truly
of additivity when the adapting field was com- independent. The adaptational abnormalities of
posed of short wavelength and long wavelength the p 1 mechanism are not predicted if the SWS
components. When the background was com- cone shows Weber adaptation preceding the sec-
posed of low luminance short wavelength and ond site. The Pugh and Mollon model can be
long wavelength components, threshold was rephrased if we consider that the SWS cone
raised more than predicted from the additivity of pathway is subject only to partial adaptation at
the component fields (Pugh, 1976). When the the first site. There is adaptation at the second
background was composed of a high luminance site that acts to subtract most of the opponent
short wavelength and lower luminance long signal.
wavelength components, threshold was raised p 4 and p 5: The remaining Stiles p mecha-
less than predicted from the additivity of the nisms are associated with detection mediated
component fields. Another irregularity of the 1 primarily by MWS and LWS cones. These mech-
to p 3 mechanisms was called the ‘limited condi- anisms obey the displacement laws and fulfil
tioning effect.’ As a long wavelength background many criteria for unitary mechanisms. The
was raised, threshold for a 435 nm test flash rose mechanism Stiles termed p 4 has often been
for about 0.7–0.8 log unit, and then stabilized, associated with the MWS photopigment and to a
i.e. showed no further effect of an increase in the lesser extent p 5 has been associated with the
adapting background. The threshold started to LWS photopigment. Nonetheless, other evidence
increase again only when the background suggested that p 4 and p 5 did not represent pho-
reached a level capable of bleaching the LWS and topigment sensitivities. This evidence came from
MWS cones. The final abnormality of the p 1 many sources: The p 4 and p 5 were broader
mechanism was one of temporal adaptation and than the spectral sensitivities of X-chromosome
termed ‘transient tritanopia.’ Following extinc- linked dichromats believed to have only one of
tion of a long wavelength background, threshold these mechanisms (Boynton, 1963). The p 4 and
for a short wavelength test flash first rose pre- p 5 mechanisms showed unexpected interac-
cipitously and then recovered very slowly. In tions that were not consistent with unitary
comparison following extinction of a short mechanisms (Boynton, Ikeda and Stiles, 1964).
wavelength background that caused the same If the parameters of the test size and duration are
steady threshold increment, sensitivity fell rap- changed, the spectral sensitivities of the isolated
idly. A model of the p 1 and p 3 pathways was mechanisms do not necessarily agree with those
suggested by Pugh and Mollon (1979) in which of the p 4 and p 5 mechanisms (Ingling and
the SWS cone was subject to two independent Martinez, 1981; Stockman and Mollon, 1986).
sources of adaptation. Adaptation could occur Finally, test sensitivity measurements yield
both in the SWS cone and at a later ‘second’ site narrower functions than the classical Stiles
where there was spectral opponency between field mechanisms (Wald, 1964; Eisner and
SWS cones and summed activity of LWS and MacLeod, 1981; Yeh et al., 1989; Stockman et al.,
MWS cones (see Chapter 6 on spectral oppo- 1993).
nency in the retina). The failure of the inde- Today it is conceded that the p mechanisms
pendence tests for p 1 was attributed to do not represent isolated cone spectral sensitiv-
independence of first site and second site adap- ities. With the acceptance of the parallel path-
tation. The limited conditioning effect of long way description of the retina (Chapter 6), it is
wavelength adapting fields was attributed to sta- recognized that the choice of spatio-temporal
bilization of the steady-state response at the sec- parameters will favor detection either in PC- or
ond site at bleaching levels of the LWS and MWS in MC-pathways. Thus, obedience of the dis-
cones. Transient tritanopia was attributed to the placement laws may represent isolation of a
adaptational properties of the second site. In higher order neural mechanism by choice of the
131
■ THE SCIENCE OF COLOR
spatio-temporal presentation, rather than isola- obtained when the background is a pedestal spa-
tion of a single photoreceptor by adaptation. tially coextensive with the test (Nacer et al.,
The spatio-temporal parameters of the stimulus 1989) and when the test is of long duration
will determine the relative sensitivities of the (King-Smith and Carden, 1976).
MC- and PC-pathway channels. If these sensi-
tivities are similar, probability summation Interpretation: The characteristic lobes and
among pathways should be considered. The notches of the increment spectral sensitivity on a
choice of a 100–200 msec square-wave pulse is white background are usually interpreted as
one that favors both MC- and PC-pathway reflecting spectrally opponent processing (see
channels. In comparison, a brief 5 msec pulse Chapter 6). Detection in the spectrally opponent
would favor the MC-pathway while a 1 second channels is most comparable to KC-pathway
gaussian-shaped pulse would favor the PC- ‘blue-on,’ and PC-pathway ‘green-on’ and ‘red-
pathway. on’ responses (Sperling and Harwerth, 1971;
Thornton and Pugh, 1983). The channels are
3.3.4 INCREMENT DETECTION ON weighted and fit to the data. This approach does
WHITE not specify adaptation in a mechanistic form, but
the weights serve this purpose. An alternative
With a white-adapting field, a test sensitivity interpretation is in terms of psychophysically
function may be obtained as a function of test specified achromatic and chromatic visual path-
wavelength for a large, long test pulse. This func- ways (King-Smith and Carden, 1976). The white
tion shows three peaks, separated by two background is considered a powerful adapting
notches (Sperling and Harwerth, 1971). There is stimulus for the achromatic but not the two
a prominent peak near 450 nm with a notch chromatic pathways. The lobes reflect weighted
near 500 nm. There are peaks at 540 and chromatic opponent activity; the notches repre-
600 nm, separated by a notch near 570 nm. The sent weighted achromatic activity. This approach
prominence of the peaks depends on the spatio- also does not specify adaptation in a mechanistic
temporal characteristics of the test stimulus and form.
the size and luminance of the background
(Figure 3.14). The most prominent peaks are
3.4 CHROMATIC
DISCRIMINATION
3.0 Discrimination refers to the ability to detect a dif-
ference between two lights that differ on some
Log Sensitivity (-Log W/m2/sr)
physical continuum. In studies of chromatic dis-
2.5
crimination, the lights may differ in wavelength,
in colorimetric purity, or in chromaticity, but do
2.0 not differ in luminance. The threshold repre-
sents some criterion change presumably corre-
lated with a difference in a neural response to
1.5
the two lights. Thresholds may be expressed in
EM (N) wavelength steps or in chromaticity steps. A
1.0 modern approach that attempts to relate detec-
tion and discrimination uses cone troland units
400 450 500 550 600 650 700 (defined in section 3.2), which allows compari-
Wavelength son of detection and discrimination in terms of
quantal excitation of the receptors. If a surround
Figure 3.14 Increment threshold spectral is present and discrimination is measured from
sensitivity function for 2 test stimuli presented on a
coextensive 4600 K, 800 td pedestal within a larger
the surround chromaticity, the threshold may
200 td surround. (Data replotted from Figure 6 of conceptually be considered as a detection, and
Miyahara et al., 1996.) related to the detection experiments described in
132
COLOR MATCHING AND COLOR DISCRIMINATION ■
section 3.3. In modern studies of chromatic dis- the observer. This technique also bridges meas-
crimination, the test chromaticity may differ urements of color matching and color discrimi-
from the surround chromaticity. nation. The standard deviation of a set of color
matches can be used as an index of color
3.4.1 HISTORICAL APPROACHES discrimination.
Figure 3.15 compares wavelength discrimina-
Wavelength discrimination: Wavelength dis- tion data from several laboratories. MacAdam’s
crimination refers to the ability of an observer to (1942) standard deviation data appear to be
detect chromatic differences along the spectrum approximately one-fifth the discrimination
locus. In the classical wavelength discrimination thresholds measured by step-by-step procedures.
experiment, the observer views a bipartite field, The peaks and valleys vary somewhat among the
one half filled with light of a standard wave- different authors. A similar variation occurs
length and the other with light of a comparison among the functions measured in different
wavelength. Both standard and comparison individuals.
fields are narrow spectral bands of light that are
varied in spectral composition and radiance. Colorimetric purity discrimination: Colori-
A common procedure is the step-by-step metric purity discrimination typically refers to
method in which both fields are initially of iden- measurements of the least colorimetric purity,
tical spectral composition and radiance (isomeric pc, the minimum amount of spectral light that
fields). The wavelength of the comparison field is allows a mixture of a spectral light and white to
changed in small steps (one nanometer or less) be distinguished from white.
and the observer adjusts the radiance of the
comparison field following each change to seek a pc Lk/(Lw Lk) (3.69)
match. Discrimination threshold is reached
when the observer reports that the fields do not where Lk is the luminance of the spectral color
appear identical regardless of the radiance of the and Lw is the luminance of the white. Figure
comparison field. The wavelength discrimination 3.16(A) shows the data expressed as the recipro-
step is expressed in terms of the difference in cal (pc) of least colorimetric purity as a function
wavelength, k, between the standard and com- of retinal illuminance. The results show a
parison fields. The procedure is repeated for a minimum in the 570–580 nm region, and
series of standard wavelengths throughout the best discrimination at the spectral extremes.
visible spectrum. Data are reported either as the
k in one direction, for example scaling toward
longer wavelengths, or as the average of k for 20
comparison lights scaled in both directions.
In an alternative technique, the standard devi-
15
ation of a repeated series of color matches is
taken as being proportional to the discrimination
step. In this procedure, the two fields are initially 10
∆λ
different in wavelength. The observer adjusts the
wavelength and radiance of the comparison field
until it appears to be the same as the standard. 5
The procedure is repeated many times, in order
to determine the standard deviation of the com-
parison field settings. In terms of experimental 0
convenience, the step-by-step method is more 400 450 500 550 600 650 700
rapid but it requires absolute calibration of Wavelength
wavelength for both the standard and the com-
Figure 3.15 Wavelength discrimination plotted as a
parison fields. The standard deviation procedure function of wavelength.The symbols show data from
is time-consuming but may be more accurate as four observers (Pokorny and Smith, 1970).The solid
it shows less dependence on criterion changes of line is the average.
133
■ THE SCIENCE OF COLOR
Colorimetric purity of a sample, S, denoted by its least amount of white added to a spectral light),
chromaticity coefficients (xS, yS) in the CIE dia- the results in the literature are rather variable
gram is related to excitation purity, pe (section but show a much flatter function than when
3.2) by: colorimetric purity is measured as the first step
from white (Jones and Lowry, 1926; Martin et
pc (yk/yS)pe (3.70) al., 1933; Wright and Pitt, 1937; Kaiser et al.,
1976). Yeh, Smith, and Pokorny (1993) reex-
where yk is the chromaticity coefficient of the amined this question carefully and found the
dominant wavelength of sample S and yS is the shape of the first step from the spectrum func-
chromaticity coefficient of the sample. tion to be highly retinal illuminance dependent
If colorimetric purity discrimination is meas- with flatter functions at higher light levels
ured as the first step from the spectrum (i.e. (Figure 3.16B).
12 12
290 TD 290 TD
11 11
10 10
9 9
110 TD 110 TD
8 8
7 7
-LOG (1-Pc)
-LOG Pc
6 6
29 TD 29 TD
5 5
4 4
3 3
2 2
2.9 TD 2.9 TD
1 1
0 0
TY TY
1 1
350 450 550 650 750 350 450 550 650 750
Wavelength Wavelength
Figure 3.16 Colorimetric purity discrimination plotted as a function of wavelength. Left panel shows
colorimetric purity thresholds measured from white and right panel shows colorimetric purity thresholds
measured from the spectrum. Data for the higher luminance levels are successively scaled vertically by 3 log
units. Data are shown for one observer, data of four other observers showed closely similar trends. (From Yeh
et al., 1993.)
134
COLOR MATCHING AND COLOR DISCRIMINATION ■
MacAdam ellipses: Wavelength and colorimet- Watson (1911) and Tyndall (1933) investi-
ric purity discrimination represent two special gated the effects of adding white light to discrim-
cases of chromaticity discrimination: discrimina- ination fields composed of spectral lights. For
tions along the spectrum locus and discrimina- wavelengths greater than 490 nm, the discrimi-
tion along axes between white and the spectrum nation step is increased with increases in the
locus. It is possible to sample chromatic discrim- white light content. Tyndall extended these
ination systematically starting at an arbitrary observations into the short wavelength region of
chromaticity (e.g. Wright, 1941). The data are the spectrum and found rather striking results.
represented in the 1931 CIE chromaticity dia- For 455 nm fields, discrimination improved with
gram (MacAdam, 1942; Brown and MacAdam, the addition of white light. Discrimination
1949; Wyszecki and Fielder, 1971). MacAdam improved and was optimal when the white light
used the standard deviation of color matches to was four to ten times higher in luminance than
represent chromaticity discrimination. For a the spectral discrimination lights. This improve-
number of different points in chromaticity space, ment has been termed the Tyndall effect. Polden
MacAdam derived a series of discrimination and Mollon (1980) coined the term ‘combinative
ellipses, which represent the discriminable dis- euchromatopsia’ to describe the enhanced
tance for a number of directions from each point. sensitivity to hue differences.
MacAdam ellipses represent data from a single
observer. Figure 3.17 shows MacAdam’s data 3.4.2 EXPERIMENTAL VARIABLES
plotted in the 1931 chromaticity diagram with
each of the ellipses representing ten times the The effects of the luminance level, spatial struc-
measured standard deviations. ture, temporal presentation and retinal position
In the equiluminant plane, discriminations are are important in chromatic discrimination. The
based solely on chromaticity differences. It is also effect of surrounds is considered separately in
possible to evaluate the joint effects of chro- section 3.4.4.
maticity and luminance in determining discrimi-
nation steps (Brown and MacAdam, 1949; Effect of field size: Chromatic discrimination
Noorlander et al., 1980). These data also describe improves when the field size is increased from 2°
ellipsoids in a three-dimensional chromaticity to 10° by a factor of about two (Brown, 1952;
and luminance space (Poirson and Wandell, Wyszecki and Stiles, 1982). The improvement is
1990). independent of the sampling direction in chro-
maticity space. Chromatic discrimination deteri-
orates when the field size is reduced below 1
530
(Pokorny and Smith, 1976). Chromatic discrimi-
520
0.8
540
nations are possible for bipartite fields as small as
510
0.7 550 3» (MacAdam, 1959). Steady fixation of small
560 fields (20» or less) leads to a deterioration of dis-
0.6 570 crimination mediated by the SWS cones (small
500
0.5 580 field tritanopia, König, 1894; Thomson and
590 Wright, 1947). The small field tritanopic effect
y
0.4 600 also occurs with steady fixation of parafoveal
610
0.3
620 fields (Hartridge, 1945; Thomson and Wright,
490 650
1947) and is reduced or eliminated by employ-
0.2 ing a scanning or glance technique (Bedford and
480
0.1 Wyszecki, 1958b; McCree, 1960).
470
460
450 380
0 Effect of retinal illuminance: Discrimination
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 functions show little change over a range of reti-
x
nal illuminance of 100–3500 troland (Bedford
Figure 3.17 The MacAdam ellipses plotted in the and Wyszecki, 1958a; Cornu and Harlay, 1969).
CIE chromaticity diagram. (From MacAdam, 1942.) However, discrimination deteriorates at low
135
■ THE SCIENCE OF COLOR
levels of retinal illuminance. The effect is partic- showed a low pass function with poor sensitivity
ularly marked for discriminations based upon to high spatial or temporal frequencies.
SWS cones (Brown, 1951; Verriest et al., 1963; Assessment of the spatial contrast sensitivity
Knoblauch et al., 1987). There is a strong inter- for equiluminant stimuli is technically demand-
action between field size and luminance; the ing due to the presence of chromatic aberration
deterioration of SWS cone discrimination is in the eye. Chromatic aberration can introduce
more pronounced with the joint reduction of unwanted luminance information in the nomi-
field size and luminance (Farnsworth, 1955; nally equiluminous grating. Since the data
Clarke, 1967; Yonemura and Kasuya, 1969). reveal that a pure color grating has low contrast
sensitivity, chromatic aberration must be care-
The gap effect: Discrimination is also affected fully minimized (Mullen, 1985) or eliminated
by the presence of a separation between two (Sekiguchi et al., 1993).
halves of a bipartite field (Sharpe and Wyszecki, A number of techniques have been used to
1976; Boynton et al., 1977). The effect of intro- assess the temporal and spatial characteristics of
ducing a gap is not the same for different types isolated receptor mechanisms. In general, the
of discriminations. Luminance discrimination is LWS and MWS mechanism have similar spatial
best when the two half fields are precisely juxta- (Brindley, 1954; Green, 1968; Cavonius and
posed. Chromatic discrimination for red–green Estévez, 1975) and temporal properties
discriminations is unimpaired by a gap and dis- (Brindley et al., 1966; Green, 1969; Estévez and
criminations based on differential SWS cone Cavonius, 1975). The SWS mechanism exhibits
excitations improve. lower spatial (Stiles, 1949; Brindley, 1954;
Green, 1968) and temporal (Brindley et al.,
Temporal presentation: Wavelength discrimi- 1966; Green, 1969; Kelly, 1974; Wisowaty and
nation improves with increasing exposure dura- Boynton, 1980) resolution.
tion but there is little agreement in the literature
as to the optimal exposure (Farnsworth, 1961; 3.4.3 MODERN APPROACH TO
Siegel, 1965; Regan and Tyler, 1971; Hita et al., CHROMATICITY DISCRIMINATION
1982). Discrimination deteriorates if stimuli are
presented successively rather than simultane- The Boynton and Kambe experiment:
ously (Uchikawa and Ikeda, 1981; Uchikawa, Boynton and Kambe (1980) performed a sys-
1983; Sachtler and Zaidi, 1992; Jin and Shevell, tematic sampling of chromaticity space in a con-
1996). stant luminance plane. Their study showed
notable differences from the earlier studies. One
Retinal location: Peripheral fixation of the major difference was that discrimination was
fields results in a marked reduction of discrimi- measured along theoretically critical axes in the
nation, particularly in the 490–530 nm equiluminant plane which correspond to the
(Moreland, 1972). Rods have been implicated as axes of the MacLeod–Boynton chromaticity
the source of degradation of discrimination space (section 3.2.5). One set of axes maintained
(Lythgoe, 1931; Stabell and Stabell, 1977; Stabell a constant amount of SWS cone stimulation,
and Stabell, 1982). s/(lm) and discrimination was measured for
test chromaticities varying in their ratios of LWS
Temporal and spatial contrast sensitivity: A to MWS cone stimulation, l/(lm). A second set
number of studies have measured the contrast of axes maintained a constant LWS to MWS cone
sensitivity function for equiluminous chromatic stimulation l/(lm) and discrimination was
alternation in space (van der Horst et al.,1967; evaluated for test chromaticities varying in SWS
Hilz and Cavonius, 1970), time (de Lange, 1958; cone excitation, s/(lm). The l/(lm) cone dis-
Kelly and van Norren, 1977; Wisowaty, 1981; criminations showed a minimum. The minimum
Swanson et al., 1987), and joint variation of occurred near a value of 0.667 for l/(lm) which
space and time (van der Horst and Bouman, is comparable to data obtained with a surround
1969; Noorlander et al., 1981). The studies report chromaticity at the equal energy spectrum (see
that equiluminous chromatic modulation section 3.4.4). The l/(lm) discriminations were
136
COLOR MATCHING AND COLOR DISCRIMINATION ■
primarily independent of the level of s/(lm), Berger, 1979). Modern studies, assessing dis-
although there appeared a small contribution of crimination on LWS, MWS cone, and on SWS
SWS cone activity at high s/(lm) levels. On the cone excitation axes have confirmed and
S cone axis, discrimination was dependent on extended this finding (Krauskopf et al., 1982;
s/(lm) at the test chromaticity. Further, Zaidi et al., 1992; Miyahara et al., 1993; Smith et
s/(lm) axis discriminations were affected only al., 2000). The data show that l/(lm) cone dis-
by the level of s/(lm) and were independent of criminations show a symmetrical V shape with
the l/(lm) component of the stimulus. minimum near the l/(lm) chromaticity of the
Subsequently, Krauskopf, Williams and Healy surround (Figure 3.18). Chromatic discrimina-
(1982) confirmed the independence of the tion data obtained with no surround show a
l/(lm) and s/(lm) lines in an experiment shallower V shape whose minimum coincides
using chromatic adaptation. They found highly with that obtained when the surround chro-
selective chromatic effects with an adaptation maticity is that of the equal energy spectrum
field that was modulated in a l/(lm) direction. (Boynton and Kambe, 1980; Yeh et al., 1993).
The l/(lm) discrimination was impaired by Discriminations on the s/(lm) axis also show a
previous adaptation to a l/(lm) stimulus, but V shape with a minimum near the s/(lm) cone
discrimination along the s/(lm) line was un- chromaticity of the surround. The V is not sym-
affected. Conversely, adaptation on a s/(lm) metrical (Zaidi et al., 1992; Miyahara et al.,
affected s/(lm) discrimination but not l/(lm) 1993), rising more sharply when the s/(lm)
discriminations. Luminance modulation had level of the test chromaticity is higher than that
little effect on chromatic thresholds. Chromatic of the surround (Figure 3.19).
modulation had little effect on luminance
thresholds. 3.4.5 INTERPRETATION
The Boynton and Kambe data when expressed
in chromaticity showed general agreement with An interpretation of chromatic discrimination
MacAdam’s (1942) ellipses. The Boynton and data at equiluminance can be viewed in terms of
Kambe experiment was designed to yield high spectral signals generated in the PC-pathway. A
criterion color-difference steps. The observer was PC-pathway retinal ganglion cell has a steady
required to identify the color direction at thresh- resting level to a steady EES adapting field of
old. Boynton and Kambe’s discrimination steps 100–1000 trolands. If the field luminance or
are equivalent to approximately 13 of MacAdam’s
standard deviations.
0.5
A second major advance in the Boynton and
Kambe formulation lies in the ability to relate
discrimination data to detection by use of the
cone troland. Boynton and Kambe showed that
Log ∆ L Trolands
0
at 115 td optimal discrimination on the l/(lm)
axis required an increase of one L td (accompa-
nied by a decrease of one M td). In comparison,
optimal discrimination on the s/(lm) axis –0.5
required an increase of eight S td.
3.4.4 THE EFFECT OF SURROUNDS –1.0
1.80 1.85 1.90 1.95 2.00
Surrounds are important since they control the Log L Trolands
state of adaptation, as was clear from detection
experiments. In early studies it was established Figure 3.18 Chromatic discrimination in the
that chromatic discrimination is best when the equiluminant plane for lights varying only in their
l/(lm) content.The observer is adapted to the equal
test chromaticity is at or near the chromaticity energy spectrum.The data are expressed in L tds (see
of the surround (Brown, 1952; Hurvich and section 3.1) with log dL plotted vs. log L at the starting
Jameson, 1961; Pointer, 1974; Loomis and chromaticity. (From unpublished data of the authors.)
137
■ THE SCIENCE OF COLOR
2.5 nation. Other visual function, including visual
acuity, is normal. Such defects represent only a
2.0 subset of the classification of human color defect
which includes color vision problems accompa-
Log ∆ S Trolands
1.5
nying hereditary and acquired eye disorders
(Pokorny et al., 1979). Congenital color defects
have fascinated visual science since their discov-
1.0
ery at the end of the eighteenth century. The fas-
cination lies in the idea that color defects
0.5 represent ‘mistakes’ of nature that can help elu-
cidate the mechanisms of normal color vision.
0 Today it is recognized that congenital color
1.00 1.50 2.00 2.50 3.00 defects arise because of point mutations,
Log S Trolands rearrangements, and deletions of the opsin genes
that determine the structure and function of the
Figure 3.19 Chromatic discrimination in the cone visual photopigments.
equiluminant plane for lights varying only in their
s/(lm) content.The observer is adapted to the equal Congenital color defects may be classified on
energy spectrum.The data are expressed in S tds (see two dimensions: a qualitative dimension which
section 3.1) with log dS plotted vs. log S at the starting states how color vision is affected and a quanti-
chromaticity. (From unpublished data of the authors.) tative dimension which describes the extent of
discrimination loss. There are three major quali-
chromaticity is changed, there is a return to near tative categories of defect, termed protan, deu-
the resting level. This is indicative of almost tan, and tritan.3 The most common congenital
complete adaptation to both chromaticity and defects are the protan and deutan defects. These
luminance, occurring before or at the generation defects show X-chromosome-linked inheritance
of the PC-pathway retinal ganglion cell signal. and occur in 8% of the European male and 0.4%
This adaptation could be either multiplicative or of the European female population (4–5% of the
subtractive, as discussed in section 3.3. If a brief total European white population). The incidence
chromatic pulse is presented, the cell responds, in the United States is considered to be higher
showing a Naka–Rushton saturation function (Paulson, 1973). The tritan defect is rarer, occur-
(Figure 3.12 in section 3.3) as the pulse is ring equally in both sexes in about 1 in 15 000 to
increased from the adaptation state. The V- 1 in 50 000 (0.002–0.007%) of the European
shapes of chromatic discrimination can be population. The tritan defect has autosomal
viewed as characteristic of saturation functions dominant inheritance. The inheritance, inci-
following an earlier stage of neural adaptation dence, and classification of these defects are
(Zaidi et al., 1992). At the point of the V, dis- summarized in Table 3.5.
crimination is assessed at the adapting chro-
maticity. These discriminations can also be 3.5.1 THE PROTAN AND DEUTAN
considered as detections (section 3.3). The DEFECTS
decrease in discrimination ability at low light
levels is consistent with the idea that the neural There are two qualitatively different forms of X-
adaptation becomes effective only above 1–10 chromosome linked congenital defect. Protan
trolands in the P pathway. observers show spectral sensitivity with long
wavelength luminosity loss. Deutan observers
show spectral sensitivity in the normal range at
3.5 CONGENITAL COLOR long wavelengths. The protan and deutan
DEFECT defects include a dichromatic form called
protanopia and deuteranopia respectively. In the
Congenital color defects represent a hereditary, dichromatic defect, the affected observer
stationary condition in which there is an abnor- requires only two primaries for full spectrum
mality of color matching and/or color discrimi- color matching. Protanopia and deuteranopia
138
COLOR MATCHING AND COLOR DISCRIMINATION ■
Table 3.5 The inheritance, incidence and classification of congenital color vision defects
Type Inheritance Incidence Classification
Protan and Deutan X-chromosome 8–10%(males)
linked recessive <1%(females)
Protanopia 1%(males) Dichromatic
Deuteranopia 1%(males) Dichromatic
Protanomaly 1%(males) Trichromatic
Deuteranomaly 5%(males) Trichromatic
Tritan defect Autosomal 0.002–0.007 Dichromatic and
dominant Trichromatic
have traditionally been regarded as a ‘reduction’
0
form of color vision (von Kries, 1897). The
dichromat was considered to lack function of 0.5
one of the normal color fundamentals. The spec- 1.0
Log Relative Sensitivity
tral sensitivity of protanopes and deuteranopes is
shown in Figure 3.20. There are also trichro- 1.5
matic forms called protanomalous trichromacy 2.0
(protanomaly) or deuteranomalous trichromacy
(deuteranomaly). The color matches of 2.5
protanomalous and deuteranomalous trichro- 3.0
mats differ from each other and from those of
3.5
normal trichromats. Anomalous trichromacy
was historically regarded as an ‘alteration’ sys- 4.0
tem (von Kries, 1897), but modern interpreta- 400 450 500 550 600 650 700
tion is in terms of inheritance of polymorphic Wavelength
forms of either the LWS (deuteranomalous) or
the MWS (protanomalous) cone photopigments Figure 3.20 Dichromatic spectral sensitivity data.
The symbols show the threshold data of Hsia and
(see section 3.2.6).
Graham.The solid lines represent the Smith and
The severity in discrimination loss of the Pokorny (1975) fundamentals of Figure 3.7, adjusted
protan and deutan defects varies considerably. vertically by eye to pass near the data.
The dichromatic form is more severe than the
trichromatic form. It should be noted that varia-
tion is minimal within a family pedigree, i.e. The assumption that protanopes and deutera-
both the qualitative and the quantitative extent nopes lack one of the normal fundamentals
of the defect are inherited. Color confusions and allows the dichromatic color matches to be rep-
spectral sensitivity of protanopes and protanom- resented in the normal chromaticity diagram.
alous trichromats or of deuteranopes and The dichromatic color matches are normalized in
deuteranomalous trichromats are qualitatively the same manner as the normal matches. The
similar, justifying the inclusive terms ‘protan’ dichromatic chromaticity coordinates fall on a
and ‘deutan’ (Farnsworth, 1947). These similar- line joining the two primaries. The chromaticity
ities also allow the design of rapid screening tests coordinates are joined to the corresponding test
for X-chromosome linked defects. wavelength, forming a confusion line. The data
Full-spectrum colorimetry, spectral sensitivity, reveal axes of discrimination loss characteristic
and wavelength and colorimetric purity discrim- of the type of defect. In the Judd revised chro-
ination functions for a limited number of maticity diagram, the confusion lines converge
observers have been described in the literature at a point, called the copunctal point (Figure
(Pokorny et al., 1979). Some major features of 3.21). The copunctal point is considered the
protan and deutan color defects are summarized locus of the ‘missing fundamental.’ It is these
in Table 3.6. copunctal points that form the new primaries in
139
■ THE SCIENCE OF COLOR
Table 3.6 Major features of congenital color vision defects
Characteristic Normal Protan Deutan Tritan
Number of 3 2(P) 2(D) 2 or 3
primaries used 3(PA) 3(DA)
in color mixture
Neutral point 494 nm 499 nm 570 nm
(dichromats only)
Copunctal point Xp = 0.7635 Xd = 1.4000 Xt = 0.1748
(dichromats only) Yp = 0.2365 Yd = 0.400 Yt = 0.0000
max of luminosity 555 nm 540 nm 560 nm 555 nm
Minimum delta 590 nm 490 nm 495 nm 590 nm
(best wavelength
discrimination)
Minimum Pc 570 nm 494 nm 499 nm 570 nm
(worst purity
discrimination)
transforming from the Judd color matching a 2 field of low photopic luminance is shown
functions to a set of König fundamentals (see in Table 3.5. This classification has been
section 3.2.5). One confusion line passes from extended (Franceschetti, 1928) and replicated
the chromaticity coordinates through the equal in large-scale studies (Schmidt, 1955; Helve,
energy spectrum to the spectrum locus. The 1972). It is important to recognize that change
spectral wavelength at the intersection is called in the parameters of study (e.g. test field size,
the neutral point. primaries, etc.) will not yield the same classifi-
The X-chromosome linked defects are diag- cation. For example, few dichromats will accept
nosed and classified using a specialized color the full range of primary ratio with an 8 test
match, the Rayleigh equation. The Rayleigh field. Most would then be classified as anom-
equation presents a match of a 589 nm spectral alous trichromats. Screening tests based upon
test light to a mixture of two spectral primaries, discrimination rather than color matching do
545 nm and 670 nm. Since the primaries fall not permit conclusive classification of color
near the linear portion of the long wavelength defect nor usually do they indicate the severity
spectrum locus, a third primary is unnecessary. of discrimination loss.
An anomaloscope is used to measure the The genetic study of the protan and deutan
Rayleigh match; the instrument is designed so defects has revealed point mutations (single
that the mixture field has fixed energy levels, nucleotide changes) or deletions within the
only the proportion of 545 nm:670 nm lights is opsin genes on the X chromosome. The opsins
varied. The 589 nm test is variable in lumi- for normal MWS and LWS photopigments lie in
nance. For some primary ratios, the trichro- a tandem array on the X chromosome. There
matic observer can adjust the 589 nm test field may be multiple copies of these genes and the
and obtain a match. Protanopic and deutera- factors controlling their expression are not yet
nopic observers can match all the primary ratios delineated. The nucleotide sequences of the
with suitable adjustment of the 589 nm test MWS and LWS opsin genes are very similar.
luminance. Normal, protanomalous and deuter- Only a few nucleotide changes differentiate
anomalous trichromats use very different whether the absorption spectrum will be that of
545 nm:670 nm primary ratios in their matches, an LWS or MWS photopigment. Some variation
thus allowing classification with high specificity. in these nucleotides (polymorphism) occurs nat-
The accepted classification of X-chromosome urally in the color-normal population. The
linked defects, based on Rayleigh matching with protan and deutan defects are correlated with
140
COLOR MATCHING AND COLOR DISCRIMINATION ■
1.0
more pronounced alterations or with deletions
Protanopic confusion lines of the opsin genes on the X-chromosome. The
qualitative classification is well correlated with
0.8 the type of gene alteration. The quantitative clas-
sification is less well correlated with the gene
array. The correlation of genotype and pheno-
0.6
type remains an area of concentrated research.
y
0.4
3.5.2 TRITAN DEFECTS
The tritan defect is characterized by a lack of
function of the mechanism that allows normals
0.2 to discriminate colors that differ by the amount
of short-wavelength light they contain. Discrimi-
nations dependent on LWS and MWS cone func-
0 tion are normal. Tritans show discrimination loss
0 0.2 0.4 0.6 0.8 1.0 for SWS-cone mediated discrimination but do
x
1.0 not show a specific alteration in color matching
that would implicate variation in the absorption
Deuteranopic confusion lines spectrum of the SWS photopigment. Some
0.8
characteristics of tritan color deficiency are
0.6 summarized in Table 3.6.
There is considerable variability in chromatic
0.4 discrimination both within and between family
pedigrees with tritan defect. Classification by
color matching is less clear-cut. Some tritans
y
0.2
make dichromatic color matches, but the major-
ity do not. Discrimination improves with
0
increase in field size (Pokorny et al., 1981).
Wright (1952) used a 1.2 field to classify and
–0.2
define tritanopia. For dichromatic tritans, the
color matches can be treated as for protanopes
–0.4
and deuteranopes. The characteristic confusion
0 0.2 0.4 0.6 0.8 1.0 1.2 1.4
axes for tritan defect are shown in Figure 3.21.
x There is no analog of anomalous trichromacy in
1.0 autosomal dominant tritan defect size (Pokorny
Tritanopic confusion lines et al., 1981).
The genetic study of the tritan defect has
0.8 revealed point mutations (single nucleotide
changes) or deletions within an opsin gene on
the 7th chromosome ascribed to the SWS pho-
0.6 topigment. The type of mutation is consistent
within a pedigree of tritan defect.
y
0.4
Figure 3.21 Dichromatic confusion lines plotted on
the Judd revised chromaticity diagram.The upper
0.2 panel shows protanopic confusion lines, the middle
panel shows deuteranopic confusion lines and the
lower panel shows tritanopic confusion lines.
0 (Copunctal points from Smith and Pokorny, 1975).
0 0.2 0.4 0.6 0.8 1.0
x
141
■ THE SCIENCE OF COLOR
3.6 ACKNOWLEDGMENT illuminance. Journal of the Optical Society of America,
48, 406–11.
Preparation work for this chapter was supported Bedford, R.E. and Wyszecki, G.W. (1958b)
in part by NIH Grant EY00901. Wavelength discrimination for point sources.
Journal of the Optical Society of America, 48, 129–35.
Bouma, P.J. (1947) Physical Aspects of Colour.
Eindhoven: N.V. Philips Gloeilampenfabrieken.
3.7 NOTES Bouman, M.A. and Koenderink, J.J. (1972)
Psychophysical basis of coincidence mechanisms in
1 The CIE has recommended standard sources for use the human visual system. Ergebnisse der Physiolgie,
in colorimetric specification. See Wyszecki and 65, 126–72.
Stiles (1982) for further information. Those men- Bowmaker, J.K., Dartnall, H.J.A., Lythgoe, J.N., and
tioned in this chapter include Standard Illuminants Mollon, J.D. (1978) The visual pigments of rods and
A, B, and C. CIE Standard Illuminant A is an incan- cones in the rhesus monkey, Macaca mulatta. Journal
descent tungsten filament lamp operated at a of Physiology (London), 274, 329–48.
color temperature of approximately 2854 K. CIE Boynton, R.M. (1963) Contributions of threshold
Standard Illuminant B approximates noon sunlight, measurements to color-discrimination theory.
with a correlated color temperature of approxi- Journal of the Optical Society of America, 53, 165–78.
mately 4870 K. CIE Standard Illuminant C approx- Boynton, R.M., Hayhoe, M.M., and MacLeod, D.I.A.
imates an overcast skylight, with a correlated color (1977) The gap effect: chromatic and achromatic
temperature of approximately 6740 K (Wyszecki visual discrimination as affected by field separation.
and Stiles, 1982). Optica Acta, 24, 159–77.
2 Differences in these equations from those in Smith Boynton, R.M., Ikeda, M., and Stiles, W.S. (1964)
and Pokorny (1975) result from rounding errors in Interactions among chromatic mechanisms as
the original calculation. Dr. Yasuhiso Nakano kindly inferred from positive and negative increment
noted and corrected them. thresholds. Vision Research, 4, 87.
3 In the older literature, these defects were termed Boynton, R.M. and Kambe, N. (1980) Chromatic dif-
‘red-blind,’ ‘green-blind,’ and ‘blue-blind.’ These ference steps of moderate size measured along the-
terms, implying failure to perceive whole domains oretically critical axes. Color Research and Application,
of color percepts, were dropped for the more agnostic 5, 13–23.
terms now used. Nonetheless, color terms such as Brindley, G.S. (1953) The effects on colour vision of
‘red–green defect’ for the protan and deutan defects adaptation to very bright lights. Journal of Physiology
and ‘blue–yellow defect’ for the tritan defect remain (London), 122, 332–50.
common. Most protans and deutans recognize a Brindley, G.S. (1954) The summation areas of human
wide variety of colors in the real world. colour-receptive mechanisms at increment thresh-
old. Journal of Physiology (London), 124, 400–8.
Brindley, G.S. (1970) Physiology of the Retina and the
Visual Pathway, 2nd edn. Baltimore, MD: Williams
3.8 REFERENCES and Wilkins.
Brindley, G.S., du Croz, J.J., and Rushton, W.A.H.
Abramov, I., Gordon, J., and Chan, H. (1991) Color (1966) The flicker fusion frequency of the blue–
appearance in the peripheral retina: effects of stim- sensitive mechanism of colour vision. Journal of
ulus size. Journal of the Optical Society of America A, 8, Physiology (London), 183, 497–500.
404–14. Brown, W.R.J. (1951) The influence of luminance
Adelson, E.H. (1982) Saturation and adaptation of the level on visual sensitivity to color differences.
rod system. Vision Research, 22, 1299–312. Journal of the Optical Society of America, 41, 684–8.
Alpern, M. (1979) Lack of uniformity in colour Brown, W.R.J. (1952) The effect of field size and chro-
matching. Journal of Physiology (London), 288, matic surroundings on color discrimination. Journal
85–105. of the Optical Society of America, 42, 837–44.
Barlow, H.B. (1956) Retinal noise and absolute Brown, W.R.J. and MacAdam, D.L. (1949) Visual sen-
threshold. Journal of the Optical Society of America, 46, sitivities to combined chromaticity and luminance
634–9. differences. Journal of the Optical Society of America,
Barlow, H.B. (1957) Increment thresholds at low 39, 808–34.
intensities considered as signal/noise discrimina- Burns, S. and Elsner, A. (1985) Color matching at
tion. Journal of Physiology, 136, 469–88. high illuminances: the color-match-area-effect and
Barlow, H.B. (1972) Dark and light adaptation: psy- photopigment bleaching. Journal of the Optical Society
chophysics. In D. Jameson and L.M. Hurvich (eds.), of America A, 2, 698–704.
Handbook of Sensory Physiology, Visual Psychophysics Burns, S.A. and Elsner, A.E. (1993) Color matching at
(Vol. VII/4, pp. 1–28). Berlin: Springer-Verlag. high illuminances: photopigment optical density
Bedford, R.E. and Wyszecki, G.W. (1958a) Luminosity and pupil entry. Journal of the Optical Society of
functions for various field sizes and levels of retinal America A, 10, 221–30.
142
COLOR MATCHING AND COLOR DISCRIMINATION ■
Cavonius, C.R. and Estévez, O. (1975) Sensitivity of Galbraith, W. and Marshall, P.N. (1985) A survey of
human color mechanisms to gratings and flicker. transforms of the CIE 1931 chromaticity diagram
Journal of the Optical Society of America, 65, 966–8. with some new non-linear transforms and histolog-
CIE (1926) Proceedings, 1924, (pp. 1–232). Cambridge: ical illustrations of their utility. Acta Histochemica, 77,
Cambridge University Press. 79–100.
CIE (1964) Proceedings, 1963 (Vienna Session), Vol. B. Geisler, W.S. (1981) Effects of bleaching and back-
(Committee Report E–1.4.1), Paris, Bureau Central grounds on the flash response of the cone system.
de la CIE, pp. 209–20. Journal of Physiology (London), 312, 413–34.
Clarke, F.J.J. (1967) Colour measurement in industry. Goldstein, E.B. and Williams, T.P. (1966) Calculated
The effect of field–element size on chromaticity dis- effects of ‘screening pigments’, Vision Research, 6,
crimination. Paper presented at the Proceedings of a 39–50.
Symposium on Colour Measurement in Industry, Grassmann, H. (1853) Zur Theorie der
London. Farbenmischung. Annalen der Physik (Leipzig), 89,
Cornsweet, T.N. (1970) Visual Perception. New York: 60–84. [English trans. Philosophical Magazine
Academic Press. (London), 1854, 7: 254–64.]
Cornu, L. and Harlay, F. (1969) Modifications de la Green, D.G. (1968) The contrast sensitivity of the
discrimination chromatique en fonction de l’eclaire- colour mechanisms of the human eye. Journal of
ment. Vision Research, 9, 1273–87. Physiology (London), 196, 415–29.
Crawford, B.H. (1949) The scotopic visibility function. Green, D.G. (1969) Sinusoidal flicker characteristics
Proceedings of the Physical Society, B, 62, 321–34. of the color-sensitive mechanisms of the eye. Vision
Crawford, B.H. (1965) Colour matching and adapta- Research, 9, 591–601.
tion. Vision Research, 5, 71–8. Grigorovici, R. and Aricescu-Savopol, I. (1958)
Dartnall, H.J.A. (1957) The Visual Pigments. London: Luminosity and chromaticity in the mesopic range.
Methuen and Company. Journal of the Optical Society of America, 48, 891–8.
de Lange, H. (1958) Research into the dynamic nature Guild, J. (1931) The colorimetric properties of the
of the human fovea–cortex systems with intermit- spectrum. Philosophical Transactions of the Royal Society
tent and modulated light. Journal of the Optical Society London A, 230, 149–87.
of America, 48, 779–89. Hammond, B.R., Wooten, B.R., and Snodderly, D.M.
Eisner, A., Fleming, S.A., Klein, M.L., and Mouldin, (1997) Individual variations in the spatial profile of
W.M. (1987) Sensitivities in older eyes with human macular pigment. Journal of Optical Society of
good acuity: Cross-sectional norms. Investigative America, A, 14, 1187–96.
Ophthalmology and Visual Science, 28, 1824–31. Hartridge, H. (1945) The change from trichromatic to
Eisner, A. and MacLeod, D.I.A. (1981) Flicker photo- dichromatic vision in the human retina. Nature,
metric study of chromatic adaptation: selective sup- 155, 657–62.
pression of cone inputs by colored backgrounds. Hayhoe, M., Benimoff, N.I., and Hood, D.C. (1987)
Journal of the Optical Society of America, 71, 705–18. The time-course of multiplicative and subtractive
Estévez, O. (1979) On the fundamental data base of adaptation processes. Vision Research, 27, 1981–96.
normal and dichromatic vision. Doctoral Hecht, S., Shlaer, S., and Pirenne, M.H. (1942) Energy,
Dissertation (Vol. 17). Amsterdam: Krips Repro quanta and vision. Journal of General Physiology, 224,
Meppel. 665–99.
Estévez, O. and Cavonius, C.R. (1975) Flicker sensi- Helve, J. (1972) A comparative study of several
tivity of the human red and green color mecha- diagnostic tests of colour vision used for measuring
nisms. Vision Research, 15, 879–81. types and degrees of congenital red–green defects.
Farnsworth, D. (1947) The Farnsworth Dichotomous Test Acta Ophthalmologica Supplement, 115, 1–64.
for Color Blindness – Panel D–15. New York: Hilz, R. and Cavonius, C.R. (1970) Wavelength
Psychological Corporation. discrimination measured with square-wave grat-
Farnsworth, D. (1955) Tritanomalous vision as a ings. Journal of the Optical Society of America, 60,
threshold function. Die Farbe, 4, 185–97. 273–7.
Farnsworth, D. (1961) A temporal factor in colour dis- Hita, E., Romero, J., Jimenez del Barco, L., and
crimination. NPL Symposium No. 8, Visual Problems of Martinez, R. (1982) Temporal aspects of color dis-
Colour (Vol. 2, pp. 65–78). New York: Chemical crimination. Journal of the Optical Society of America,
Publishing Co. 72, 578–82.
Franceschetti, A. (1928) Die Bedeutung der Hood, D.C. (1998) Lower-level visual processing and
Einstellungsbreite am Anomaloskop fur die Diagnose models of light adaptation. Annual Review of
der eizelnen Typen der Farbensinnstorungen Psychology, 49, 503–35.
nebst Bemerkungen uber ihre Vererbungsmodus. Hood, D.C. and Finkelstein, M.A. (1986) Sensitivity to
Schweizerische Medizinische Wochenschrift, 52, 1273–8. light. In K.R. Boff, L. Kaufman, and J.P. Thomas
Fridrikh, L. (1957) Colour-combination curves for (eds), Handbook of Perception and Human Performance,
normal trichromats determined by direct energy Vol I: Sensory Processes and Perception, pp. 5-1–5-66).
measurements. Biophysics [translation of Biofizika], New York: John Wiley and Sons.
2, 129–32. Horner, R.G. and Purslow, E.T. (1947) Dependence of
143
■ THE SCIENCE OF COLOR
anomaloscope matches on viewing distance or field empfindungen und ihre Intensitats–Vertheilung im
size. Nature, 160, 23–4. Spectrum. Akademie der Wissenschaften Berlin,
Hurvich, L.M. and Jameson, D. (1961) Opponent Sitzungsberichte, Pt.2, pp. 805–29.
chromatic induction and wavelength discrimina- König, A. and Dieterici, C. (1893) Die Grund-
tion. In R. Jung and H. Kornhuber (eds), The Visual empfindungen in normalen und anomalen Farben
System: Neurophysiology and Psychophysics. Berlin: Systemen und ihre Intensitats-Verthielung im
Springer, pp. 144–52. Spectrum. Zeitschrift für Psychologie und Physiologie der
Ingling, C.R. and Martinez, E. (1981) Stiles’ P 5 mech- Sinnesorgane, 4, 241–347.
anism: failure to show univariance is caused by Krauskopf, J., Williams, D.R., and Heeley, D.W. (1982)
opponent–channel input. Journal of the Optical Society Cardinal directions of color space. Vision Research,
of America, 71, 1134–7. 22, 1123–31.
Jin, E.W. and Shevell, S.K. (1996) Color memory and Le Grand, Y. (1968) Light, Colour and Vision, 2nd edn.
color constancy. Journal of the Optical Society of London: Chapman and Hall, pp. 1–564.
America A, 13, 1981–91. Lee, B.B., Pokorny, J., Smith, V.C., Martin, P.R., and
Jones, L.A. and Lowry, E.M. (1926) Retinal sensibility Valberg, A. (1990) Luminance and chromatic mod-
to saturation differences. Journal of the Optical Society ulation sensitivity of macaque ganglion cells and
of America, 13, 25–34. human observers. Journal of the Optical Society of
Judd, D.B. (1930) Reduction of data on mixture of America A, 7, 2223–36.
color stimuli. Journal of Research National Bureau of Loomis, J.M. and Berger, T. (1979) Effects of chro-
Standards (USA), 4, 515–48. matic adaptation on color discrimination and color
Judd, D.B. (1935) A Maxwell triangle yielding uni- appearance. Vision Research, 19, 891–901.
form chromaticity scales. Journal of the Optical Society Lythgoe, R. (1931) Dark-adaptation and the periph-
of America, 25, 24–35. eral colour sensations of normal subjects. British
Judd, D.B. (1951a) Basic correlates of the visual stim- Journal of Ophthalmology, 15, 193–210.
ulus. In S.S. Stevens (ed.), Handbook of Experimental MacAdam, D.L. (1942) Visual sensitivities to color dif-
Psychology. New York: Wiley, pp. 811–67. ferences in daylight. Journal of the Optical Society of
Judd, D.B. (1951b) Colorimetry and artificial daylight. America, 32, 247–74.
In Technical Committee No. 7 Report of Secretariat MacAdam, D.L. (1959) Small-field chromaticity dis-
United States Commission, International crimination. Journal of the Optical Society of America,
Commission on Illumination, Twelfth Session, 49, 1143–6.
Stockholm, pp. 1–60. MacLeod, D.I.A. and Boynton, R.M. (1979)
Kaiser, P.K., Comerford, J.P., and Bodinger, D.M. Chromaticity diagram showing cone excitation by
(1976) Saturation of spectral lights. Journal of the stimuli of equal luminance. Journal of the Optical
Optical Society of America, 66, 818–26. Society of America, 69, 1183–5.
Kelly, D.H. (1974) Spatio-temporal frequency charac- Martin, L.C., Warburton, F.L., and Morgan, W.J.
teristics of color-vision mechanisms. Journal of the (1933) The determination of the sensitiveness of the
Optical Society of America, 64, 983–90. eye to differences in the saturation of colours. Great
Kelly, D.H. and van Norren, D. (1977) Two-band Britain Medical Research Council, 188, 5–42.
model of heterochromatic flicker. Journal of the Maxwell, J.C. (1860) On the theory of compound col-
Optical Society of America, 67, 1081–91. ors and relations of the colors of the spectrum.
King-Smith, P.E. and Carden, D. (1976) Luminance Philosophical Transactions, 150, 57–84. [Reprinted
and opponent-color contributions to visual detec- with commentary by Qasim Zaidi in: Color Research
tion and adaptation and to temporal and spatial and Application, 1993, 18, 270–87.]
integration. Journal of the Optical Society of America, McCree, K.J. (1960) Small-field tritanopia and the
66, 709–17. effects of voluntary fixation. Optica Acta, 7, 317–23.
King-Smith, P.E. and Webb, J.R. (1974) The use of Merbs, S.L. and Nathans, J. (1992) Absorption spectra
photopic saturation in determining the fundamen- of human cone pigments. Nature, 356, 433–5.
tal spectral sensitivity curves. Vision Research, 14, Miles, W.R. (1954) Comparison of the functional and
421–9. structural areas in human fovea. I. Method of
Knoblauch, K., Saunders, F., Kusuda, M., Hynes, R., entoptic plotting. Journal of Neurophysiology, 17,
Podgor, M., Higgens, K.E., and deMosasterio, M. 22–38.
(1987) Age and illuminance effects in the Miller, S.S. (1972) Psychophysical estimates of visual
Farnsworth–Munsell 100-hue test. Applied Optics, pigment densities in red–green dichromats. Journal
26, 1441–8. of Physiology (London), 223, 89–107.
Koenderink, J.J., van de Grind, W.A., and Bouman, Miyahara, E., Pokorny, J., and Smith, V.C. (1996)
M.A. (1970) Models of retinal signal processing at Increment threshold and purity discrimination
high luminances. Kybernetik, 6, 227–37. spectral sensitivities of X-chromosome-linked color
König, A. (1894) Uber den menschlichen Sehpurpur defective observers. Vision Research, 36, 1597–613.
und seine Bedeutung fur das Sehen. Akademie der Miyahara, E., Smith, V.C., and Pokorny, J. (1993)
Wissenschaften Berlin, Sitzungsberichte, pp. 577–98. How surrounds affect chromaticity discrimination.
König, A. and Dieterici, C. (1886) Die Grund- Journal of the Optical Society of America A, 10, 545–53.
144
COLOR MATCHING AND COLOR DISCRIMINATION ■
Mollon, J.D. and Polden, P.G. (1977) An anomaly in Pokorny, J. and Smith, V.C. (1981) A variant of
the response of the eye to light of short wave- red–green color defect. Vision Research, 21, 311–17.
lengths. Philosophical Transactions of the Royal Society of Pokorny, J. and Smith, V.C. (1986) Colorimetry and
London – Series B: Biological Sciences, 278, 207–40. Color Discrimination. In K.R. Boff, L. Kaufman, and
Moreland, J.D. (1972) Peripheral Colour Vision. In D. J P. Thomas (eds), Handbook of Perception and Human
Jameson and L.M. Hurvich (eds), Handbook of Performance, Vol I: Sensory Processes and Perception.)
Sensory Physiology, Visual Psychophysics, Vol. VII/4. New York: John Wiley and Sons, pp. 8-1–8-51.
Berlin: Springer-Verlag, pp. 517–36. Pokorny, J. and Smith, V.C. (1997) How much light
Moreland, J.D. and Bhatt, P. (1984) Retinal distribu- reaches the retina? In C.R. Cavonius (ed.), Colour
tion of macular pigment. Documenta Ophthalmologica Vision Deficiencies XIII. Documenta Ophthalmologica
Proceedings Series, 39, 127–32. Proceedings Series, 59, 491–511.
Moreland, J.D. and Cruz, A. (1959) Colour perception Pokorny, J., Smith, V. C., and Lutze, M. (1987) Aging
with the peripheral retina. Optica Acta, 6, 117–51. of the human lens. Applied Optics, 26, 1437–40.
Moreland, J.D., Torczynski, E., and Tripathi, R. (1991) Pokorny, J., Smith, V.C., and Starr, S.J. (1976)
Rayleigh and Moreland matches in the ageing eye. Variability of color mixture data – II. The effect of
Documenta Ophthalmologica Proceedings Series, 54, viewing field size on the unit coordinates. Vision
347–52. Research, 16, 1095–8.
Mullen, K.T. (1985) The contrast sensitivity of human Pokorny, J., Smith, V.C., Verriest, G., and Pinckers,
colour vision to red–green and blue–yellow chro- A.J.L.G. (eds) (1979) Congenital and Acquired Color
matic gratings. Journal of Physiology (London), 359, Vision Defects. New York: Grune and Stratton.
381–400. Pokorny, J., Smith, V.C., and Went, L.N. (1981)
Nacer, A., Murray, I.J., and Carden, D. (1989) Color matching in autosomal dominant tritan
Interactions between luminance mechanisms and defect. Journal of the Optical Society of America, 71,
colour opponency. In C.M. Dickinson and I.J. 1327–34.
Murray (eds), Seeing Contour and Colour. Oxford: Polden, P.G. and Mollon, J.D. (1980) Reversed effects
Pergamon Press, pp. 357–60. of adapting stimuli on visual sensitivity. Proceedings
Neitz, M., Neitz, J., and Jacobs, G.H. (1991) Spectral of the Royal Society of London Series B, 210, 235–72.
tuning of pigments underlying red–green color Polyak, S.L. (1957) The Vertebrate Visual System.
vision. Science, 252, 971–3. Chicago: University of Chicago Press.
Noorlander, C., Heuts, M.J.G., and Koenderink, J.J. Pugh, E.N. (1976) The nature of the p-1 colour mech-
(1980) Influence of the target size on the detection anism of W.S. Stiles. Journal of Physiology (London),
threshold for luminance and chromaticity contrast. 257, 713–47.
Journal of the Optical Society of America, 70, 1116–21. Pugh, E.N.J. and Kirk, D.B. (1986) The pi mechanisms
Noorlander, C., Heuts, M.J.G., and Koenderink, J.J. of WS Stiles: an historical review. [Review].
(1981) Sensitivity to spatiotemporal combined Perception, 15, 705–28.
luminance and chromaticity contrast. Journal of the Pugh, E.N.J. and Mollon, J.D. (1979) A theory of the
Optical Society of America, 71, 453–9. p-1 and p-3 color mechanisms of Stiles. Vision
Palmer, D.A. (1978) Maxwell spot and additivity in Research, 19, 293–312.
tetrachromatic matches. Journal of the Optical Society Regan, D. and Tyler, C.W. (1971) Temporal summa-
of America, 68, 1501–5. tion and its limit for wavelength changes: an analog
Palmer, D.A. (1981) Nonadditivity in color matches of Bloch’s Law for color vision. Journal of the Optical
with four instrumental primaries. Journal of the Society of America, 61, 1414–21.
Optical Society of America, 71, 966–9. Richards, W. and Luria, S.M. (1964) Color-mixture
Paulson, H.M. (1973) Comparison of color vision tests functions at low luminance levels. Vision Research, 4,
used by the Armed Forces. Color Vision. National 281–313.
Academy of Sciences. Ruddock, K.H. (1963) Evidence for macular pigmen-
Pease, P.L., Adams, A.J., and Nuccio, E. (1987) Optical tation from colour matching data. Vision Research, 3,
density of human macular pigment. Vision Research, 417–29.
27, 705–10. Rushton, W.A.H. (1972) Visual pigments in man. In
Pointer, M.R. (1974) Color discrimination as a func- H.J.A. Dartnall (ed.), Handbook of Sensory Physiology,
tion of observer adaptation. Journal of the Optical Vol. VII/I. Berlin: Springer, pp. 364–94.
Society of America, 64, 750–9. Sachtler, W.L. and Zaidi, Q. (1992) Chromatic and
Poirson, A.B. and Wandell, B.A. (1990) The ellipsoidal luminance signals in visual memory. Journal of the
representation of spectral sensitivity. Vision Research, Optical Society of America A, 9, 877–94.
30, 647–52. Sanocki, E., Shevell, S.K., and Winderickx, J. (1994)
Pokorny, J. and Smith, V.C. (1970) Wavelength discrim- Serine/Alanine amino acid polymorphism of the L-
ination in the presence of added chromatic fields. cone photopigment assessed by dual Rayleigh-type
Journal of the Optical Society of America, 69, 562–9. color matches. Vision Research, 34, 377–82.
Pokorny, J. and Smith, V. C. (1976) Effect of field size Schmidt, I. (1955) Some problems related to testing
on red–green color mixture equations. Journal of the color vision with the Nagel anomaloscope. Journal of
Optical Society of America, 66, 705–8. the Optical Society of America, 45, 514–22.
145
■ THE SCIENCE OF COLOR
Schrödinger, E. (1920) Grundlinien einer Theorie der Physiology and Psychophysics. London: Academic
Farbenmetrik im Tagessehen. Annalen der Physik Press, pp. 93–105.
(Leipzig), 63, 134–82. English translation in D.L. Speranskaya, N.I. (1959) Determination of spectrum
MacAdam (ed.), Sources of Color Science. Cambridge, color coordinates for twenty-seven normal
MA: The MIT Press, 1970. observers. Optics and Spectroscopy, 7, 424–8.
Schrödinger, E. (1925) Ueber das Verhaltnis der Speranskaya, N.I. (1961) Methods of determination
Vierfarben- zur Dreifarbentheorie. Sitzungsberichte of the co-ordinates of spectrum colours. NPL
der Mathematisch–naturwissenschaftlichen Klasse der Symposium No. 8, Visual Problems of Colour (Vol. 1, pp.
Kaiserlichen Akademie der Wissenschaften Wien, 134, 319–325). New York: Chemical Publishing Co.
Abt. IIa, 471–90. [Commentary by Qasim Zaidi and Sperling, H.G. and Harwerth, R.S. (1971) Red–green
National Translation Center translation ‘On the cone interaction in the increment-threshold spec-
relationship of four-color theory to three-color tral sensitivity of primates. Science, 172, 180–4.
theory’, Color Research and Application, 1994, 19, Stabell, B. and Stabell, U. (1976) Rod and cone con-
37–47.] tributions to peripheral colour vision. Vision
Sekiguchi, N., Williams, D.R., and Brainard, D.H. Research, 16, 1099–104.
(1993) Aberration-free measurements of the visibil- Stabell, U. and Stabell, B. (1977) Wavelength discrim-
ity of isoluminant gratings. Journal of the Optical ination of peripheral cones and its change with rod
Society of America A, 10, 2105–17. intrusion. Vision Research, 17, 423–6.
Shapiro, A.G., Pokorny, J., and Smith, V.C. (1994) Stabell, U. and Stabell, B. (1980) Variation in density
Rod contribution to large field color-matching. Color of macular pigmentation and in short-wave cone
Research and Application, 19, 236–45. sensitivity with eccentricity. Journal of the Optical
Shapiro, A.G., Pokorny, J., and Smith, V.C. (1996) Society of America, 70, 706–11.
Cone-rod receptor spaces, with illustrations that use Stabell, U. and Stabell, B. (1982) Color vision in the
CRT phosphor and light-emitting-diode spectra. peripheral retina under photopic conditions. Vision
Journal of the Optical Society of America A, 13, Research, 22, 839–44.
2319–28. Stiles, W.S. (1949) Increment thresholds and the
Sharpe, L.T. and Wyszecki, G. (1976) Proximity factor mechanisms of colour vision. Documenta
in color-difference evaluations. Journal of the Optical Ophthalmologica, 3, 138–63.
Society of America, 66, 40–9. Stiles, W.S. (1955) 18th Thomas Young Oration. The
Shevell, S.K. (1977) Saturation in human cones. basic data of colour-matching. The Yearbook of the
Vision Research, 17, 427–34. Physical Society, pp. 44–65.
Siegel, M.H. (1965) Color discrimination as a function Stiles, W.S. (1959) Color vision: The approach
of exposure time. Journal of the Optical Society of through increment threshold sensitivity. Proceedings
America, 55, 566–8. National Academy of Sciences USA, 45, 100–14.
Smith, V. and Pokorny, J. (1973) Psychophysical esti- Stiles, W.S. (1978) Mechanisms of Colour Vision. London:
mates of optical density in human cones. Vision Academic Press.
Research, 13, 1099–202. Stiles, W.S. and Burch, J.M. (1955) Interim report to
Smith, V.C. and Pokorny, J. (1975) Spectral sensitivity the Commission Internationale de l’Eclairage,
of the foveal cone photopigments between 400 and Zurich, 1955, on the National Physical Laboratory’s
500 nm. Vision Research, 15, 161–71. investigation of colour-matching. Optica Acta, 2,
Smith, V.C. and Pokorny, J. (1977) Large-field trichro- 168–81.
macy in protanopes and deuteranopes. Journal of the Stiles, W.S. and Burch, J.M. (1959) NPL colour-
Optical Society of America, 67, 213–20. matching investigation: final report. Optica Acta, 6,
Smith, V.C. and Pokorny, J. (1996) The design and use 1–26.
of a cone chromaticity space. Color Research and Stockman, A., MacLeod, D.M., and Johnson, N.E.
Application, 21, 375–83. (1993) Spectral sensitivities of the human cones.
Smith, V.C., Pokorny, J., Lee, B.B., and Dacey, D.M. Journal of the Optical Society of America, 10, 2491–521.
(2001) Primate Horizontal Cell Dynamics: An Stockman, A., MacLeod, D.M., and Vivien, J.A. (1993)
Analysis of Sensitivity Regulation in the Outer Isolation of the middle- and long-wavelength sensi-
Retina. Journal of Neurophysiology, 85, 545–58. tive cones in normal trichromats. Journal of the
Smith, V.C., Pokorny, J., and Starr, S.J. (1976) Optical Society of America, 10, 2471–90.
Variability of color mixture data – I. Interobserver Stockman, A. and Mollon, J. (1986) The spectral sen-
variability in the unit coordinates. Vision Research, sitivities of the middle- and long-wavelength cones:
16, 1087–94. an extension of the two-colour threshold technique
Smith, V.C., Pokorny, J., and Sun, H. (2000) of W.S. Stiles. Perception, 15, 729–54.
Chromatic contrast discrimination: Data and predic- Stockman, A. and Sharpe, L.T. (2000) The spectral
tion for stimuli varying in L and M cone excitation. sensitivities of the middle- and long-wavelength-
Color Research and Application, 25, 105–15. sensitive cones derived from measurements in
Smith, V.C., Pokorny, J., and Zaidi, Q. (1983) How do observers of known genotype. Vision Research, 40,
sets of color-matching functions differ? In J.D. 1711–37.
Mollon and L.T. Sharpe (eds), Colour Vision: Swanson, W.H. and Fish, G.E. (1996) Age-related
146
COLOR MATCHING AND COLOR DISCRIMINATION ■
changes in the color-match-area effect. Vision absorption in the visible and its consequences for
Research, 36, 2079–85. the cone receptor primaries (2F1972–17). TNO
Swanson, W.H., Ueno, T., Smith, V.C., and Pokorny, J. Report, Institute for Perception, Soesterberg The
(1987) Temporal modulation sensitivity and pulse Netherlands.
detection thresholds for chromatic and luminance Vos, J.J. (1978) Colorimetric and photometric proper-
perturbations. Journal of the Optical Society of America ties of a 2 fundamental observer. Color Research and
A, 4, 1992–2005. Application, 3, 125–8.
Tan, K.E.W.P. (1971) Vision in the Ultraviolet. Utrecht: Vos, J.J. and Walraven, P.L. (1971) On the derivation
Drukkerij Elinkwijk. of the foveal receptor primaries. Vision Research, 11,
Terstiege, H. (1967) Untersuchungen zum Persistenz- 799–818.
und Koeffizientensatz. Die Farbe, 16, 1–120. Wald, G. (1964) The receptors of human color vision.
Thomson, L.C. and Wright, W.D. (1947) The Science, 145, 1007.
colour sensitivity of the retina within the central Watson, W. (1911) Note on the sensibility of the eye
fovea of man. Journal of Physiology (London), 105, to variations of wave-length. Proceedings of the Royal
316–31. Society (London), B84, 118–21.
Thornton, J.E. and Pugh, E.N.J. (1983) Red/green Weale, R.A. (1988) Age and the transmittance of the
color opponency at detection threshold. Science, human crystalline lens. Journal of Physiology
219, 191–3. (London), 395, 577–87.
Trezona, P.W. (1970) Rod participation in the ‘blue’ Werner, J.S., Donnelly, S.K., and Kliegl, R. (1987)
mechanism and its effect on colour matching. Vision Aging and human macular pigment density. Vision
Research, 10, 317–32. Research, 27, 257–68.
Troland, L.T. (1915) The theory and practise of the Wisowaty, J.J. (1981) Estimates for the temporal
artificial pupil. Psychological Review, 22, 167–76. response characteristics of chromatic pathways.
Troy, J.B. and Lee, B.B. (1994) Steady discharges of Journal of the Optical Society of America, 71, 970–7.
macaque retinal ganglion cells. Visual Neuroscience, Wisowaty, J.J. and Boynton, R.M. (1980) Temporal
11, 111–18. modulation sensitivity of the blue mechanism:
Troy, J.B. and Robson, J.G. (1992) Steady discharges measurements made without chromatic adaptation.
of X and Y retinal ganglion cells of cat under pho- Vision Research, 20, 895–909.
topic illuminance. Visual Neuroscience, 9, 535–53. Wright, W.D. (1929) A re-determination of the
Tyndall, E.P.T. (1933) Chromaticity sensibility to trichromatic coefficients of the spectral colours.
wave-length difference as a function of purity. Transactions of the Optical Society, 30, 141–64.
Journal of the Optical Society of America, 23, 15–24. Wright, W.D. (1941) The sensitivity of the eye to small
Uchikawa, K. (1983) Purity discrimination: successive colour differences. Proceedings of the Physical Society
vs simultaneous comparison method. Vision (London), 53, 93–112.
Research, 23, 53–8. Wright, W.D. (1946) Researches on Normal and Defective
Uchikawa, K. and Ikeda, M. (1981) Temporal deterio- Colour Vision. London: Henry Kimpton.
ration of wavelength discrimination with successive Wright, W.D. (1952) The characteristics of tritanopia.
comparison method. Vision Research, 21, 591–5. Journal of the Optical Society of America, 42, 509–20.
van der Horst, G.J.C. and Bouman, M.A. (1969) Wright, W.D. and Pitt, F.H.G. (1937) The saturation-
Spatiotemporal chromaticity discrimination. Journal discrimination of two trichromats. Proceedings of the
of the Optical Society of America, 59, 1482–8. Physical Society (London), 49, 329–31.
van der Horst, G.J.C., de Weert, C.M.M., and Wyszecki, G. and Fielder, G.H. (1971) New color-
Bouman, M.A. (1967) Transfer of spatial chromatic- matching ellipses. Journal of the Optical Society of
ity contrast at threshold in the human eye. Journal America, 61, 1135–52.
of the Optical Society of America, 57, 1260–7. Wyszecki, G. and Stiles, W.S. (1980) High-level
van Norren, D. and Vos, J.J. (1974) Spectral transmis- trichromatic color matching and the pigment-
sion of the human ocular media. Vision Research, 14, bleaching hypothesis. Vision Research, 20, 23–37.
1237–44. Wyszecki, G. and Stiles, W.S. (1982) Color Science –
Verriest, G., Buyssens, A., and Vanderdonck, R. (1963) Concepts and Methods, Quantitative Data and Formulae,
Etude quantitative de l’effet qu’exerce sur les resul- 2nd edn. New York: John Wiley and Sons.
tats de quelques tests de la discrimination chroma- Yeh, T., Pokorny, J., and Smith, V.C. (1993) Chromatic
tique une diminution non selective du niveau d’un discrimination with variation in chromaticity and
eclairage. Revue d’Optique, 3, 105–19. luminance: data and theory. Vision Research, 33,
Viénot, F. (1983) Can variation in macular pigment 1835–45.
account for the variation of colour matches with Yeh, T., Smith, V.C., and Pokorny, J. (1989) The effect
retinal position? In J.D. Mollon and L.T. Sharpe of background luminance on cone sensitivity func-
(eds), Colour Vision. New York: Academic Press, pp. tions. Investigative Ophthalmology and Visual Science,
107–16. 30, 2077–86.
von Kries, J. (1897) Uber Farbensysteme. Zeitschrift für Yeh, T., Smith, V. C., and Pokorny, J. (1993)
Psychologie Physiologie Sinnesorg, 13, 241–324. Colorimetric purity discrimination: data and theory.
Vos, J.J. (1972) Literature review of human macular Vision Research, 33, 1847–57.
147
■ THE SCIENCE OF COLOR
Yonemura, G.T. and Kasuya, M. (1969) Color discrim- Zaidi, Q. (1986) Adaptation and color matching. Vision
ination under reduced angular subtense and lumi- Research, 26, 1925–38.
nance. Journal of the Optical Society of America, 59, Zaidi, Q., Shapiro, A., and Hood, D. (1992) The effect
131–5. of adaptation on the differential sensitivity of the S-
Young, T. (1802) On the theory of light and colours. cone color system. Vision Research, 32, 1297–318.
Philisophical Transactions (London), 92, 12–48.
148
View publication stats
You might also like
- Encyclopedia of Physical Science and Technology 3e Measurements Techniques and Instrumentation PDFNo ratings yetEncyclopedia of Physical Science and Technology 3e Measurements Techniques and Instrumentation PDF368 pages
- Algorithms and Applications in Computer Vision: Lihi Zelnik-ManorNo ratings yetAlgorithms and Applications in Computer Vision: Lihi Zelnik-Manor88 pages
- Digital Image Representation & Color VisionNo ratings yetDigital Image Representation & Color Vision52 pages
- Color Science: Concepts and Methods, Quantitative Data and Formulae0% (3)Color Science: Concepts and Methods, Quantitative Data and Formulae8 pages
- Light and The EM Spectrum - The H.V.S. and Color PerceptionNo ratings yetLight and The EM Spectrum - The H.V.S. and Color Perception75 pages
- SC076 Understanding Color: Giordano Beretta Hewlett-Packard CoNo ratings yetSC076 Understanding Color: Giordano Beretta Hewlett-Packard Co90 pages
- Macadam1942 - Visual Sensitive To Color Diferences in DaylightNo ratings yetMacadam1942 - Visual Sensitive To Color Diferences in Daylight28 pages
- Color Spaces and Digital Imaging (PDFDrive)No ratings yetColor Spaces and Digital Imaging (PDFDrive)7 pages
- Predicting and Specifying The Perceived Colors of Reflective ObjectsNo ratings yetPredicting and Specifying The Perceived Colors of Reflective Objects10 pages
- K - Artigo - Review of The CIE Color System and Its Use in Dentisty PDFNo ratings yetK - Artigo - Review of The CIE Color System and Its Use in Dentisty PDF9 pages
- Color Vision and Colorimetry - Theory and ApplicationsNo ratings yetColor Vision and Colorimetry - Theory and Applications188 pages
- Color Imaging Fundamentals and ApplicationsNo ratings yetColor Imaging Fundamentals and Applications1,041 pages
- Color Imaging Fundamentals and Applications PDFNo ratings yetColor Imaging Fundamentals and Applications PDF1,041 pages
- 1980 - Buchsbaum. A Spatial Processor Model For Object Colour PerceptionNo ratings yet1980 - Buchsbaum. A Spatial Processor Model For Object Colour Perception26 pages
- Colour Specification & Measurement: Dr. Neha Mehra100% (1)Colour Specification & Measurement: Dr. Neha Mehra15 pages
- Introduction To Color Imaging Science Hsien-Che Lee PDF DownloadNo ratings yetIntroduction To Color Imaging Science Hsien-Che Lee PDF Download75 pages
- The Color Metamerism Evaluation of Paint Based OnNo ratings yetThe Color Metamerism Evaluation of Paint Based On7 pages
- Colorimetry & Color Specification SystemNo ratings yetColorimetry & Color Specification System15 pages
- Color - Color in The Eye: Rods vs. ConesNo ratings yetColor - Color in The Eye: Rods vs. Cones5 pages
- Color Order System and Color Matching Functions2016 and Activity 12No ratings yetColor Order System and Color Matching Functions2016 and Activity 1227 pages
- Datasheet STW8C12C-E0 (R9+) R2.4 240607No ratings yetDatasheet STW8C12C-E0 (R9+) R2.4 24060727 pages
- Principles Governing Photometry 2nd Edition: Bureau International Des Poids Et MesuresNo ratings yetPrinciples Governing Photometry 2nd Edition: Bureau International Des Poids Et Mesures44 pages
- 1977 (Colorimetria e Coloranti) - Studies in Conservation - 22 - 161-169No ratings yet1977 (Colorimetria e Coloranti) - Studies in Conservation - 22 - 161-16910 pages
- Utilization of ITA and Hue Angle in The Measurement of Skin Color ImagesNo ratings yetUtilization of ITA and Hue Angle in The Measurement of Skin Color Images8 pages
- (Sensor de Color) Aplication Note - VEML6040No ratings yet(Sensor de Color) Aplication Note - VEML604015 pages
- Jaundice Eye Color Index (JECI) : Quantifying The Yellowness of The Sclera in Jaundiced Neonates With Digital PhotographyNo ratings yetJaundice Eye Color Index (JECI) : Quantifying The Yellowness of The Sclera in Jaundiced Neonates With Digital Photography7 pages