1/12/2009
PAM3012 Digital Image Processing for Radiographers Image S I Sampling & Q li Quantization ti ti
In this lecture
Definitions of Sampling & Quantisation Representing Digital Images Spatial & G l Gray-l Gray-level resolution l l Perceived Image Quality & Resolution Aliasing & Moire patterns
Sampling & Quantisation
The spatial and amplitude digitization of
f(x,y) is called:
image sampling when it refers to spatial coordinates (x,y) and gray-level quantization when it refers to graythe amplitude.
Sampling & Quantisation
1D function is a plot of amplitude values of the continuous image along AB
Sampling & Quantisation
Sampling limitations set by detector Single Sensor Line Sensor Array
Mechanical motion in x and y can be very precise Limited by focussing optics Number of elements in strip Density of elements in array
Sampling & Quantisation
Continuous image projected onto Array Sensor
1/12/2009
Representing Digital Images
Sampling & quantisation produces a
matrix of numbers
Representing Digital Images
Coordinate convention used to represent digital images (0, 1) second sample in first row NOT the values of physical coordinates when image was sampled
Image f(x,y) is sampled producing digital
image with M rows and N columns quantities
Values of x & y become discrete
Representing Digital Images
M X N image can be written in matrix from:
f (0,0) f (1,0) f ( x, y ) = ... f ( N 1,0) f (0,1) ... ... ... f (0, M 1) f (1, M 1) ... f ( N 1, M 1)
Representing Digital Images
Can also be written in traditional matrix form:
a 0, 0 a 1, 0 A= ... aN 1, 0
a 0,1
... ...
... ... f ( N 1,1) ...
aN 1 , 1
... a0,M 1 1 ... a1,M 1 ... ... ... aN 1,M 1
Digital Image
Image Elements (Pixels)
Digital Image
Image Elements (Pixels)
aij = f (x = i , y = j ) = f (i , j )
Representing Digital Images
f(x,y) is a digital image if:
(x,y) are integers f is a function that assigns a gray-level grayvalue (from R) to each distinct pair of coordinates (x,y) [quantization]
Representing Digital Images
The digitization process requires decisions
about:
values for N, M Where N x M: the image array and the number of discrete gray levels allowed for each pixel.
Gray levels are usually integers
1/12/2009
Representing Digital Images
These quantities are usually integer powers
of two:
N=2n M=2m and G=2k
Sampling & Quantization
Calculating number of storage bits
number of bits = N x M x k
N&M = k = number of samples in vertical & horizontal direction bit depth
G = number of gray levels
Another assumption is that the discrete
levels are equally spaced between 0 and L-1 Lin the gray scale.
Greater number of bits => higher image quality
Spatial Resolution
Smallest discernible detail in an image Sampling is the principle factor Often defined by line pairs per mm
GrayGray-Level Resolution
Smallest discernible
change in gray-level, L gray-
Number of gray-levels gray Determined by ADC
Width of line, W Width of line pair, 2W 1/2W line pairs per mm
Examples
Reducing spatial resolution, constant Gray-level Gray 1024 X 1024 image subsampled down to 32 X 32 8-bit image 256 gray levels
Examples
Reducing spatial resolution, constant Gray-level Gray 1024 X 1024 image subsampled down to 32 X 32 8-bit image 256 gray levels
1/12/2009
Examples
Reducing Gray-level, constant spatial resolution (452 X 374) Gray 8-bit (256 levels) to 1-bit (2 levels) 1-
Sampling & Quantization
How many samples and gray levels are
required for a good approximation?
Quality of an image depends on number of pixels and gray-level number gray i.e. the more these parameters are increased, the closer the digitized array approximates the original image. But: Storage & processing requirements increase rapidly as a function of N, M, and k
Perceived Image Quality & Resolution
Different versions (images) of the same
object can be generated through:
Varying N, M numbers N Varying k (number of bits) Varying both
Isopreference Curves
Isopreference curves (in the N-k plane) N Each point: image having values of N and k equal to the coordinates of thi point l t th di t f this i t Points lying on an isopreference curve correspond to images of equal subjective quality.
Huang 1965 Isopreference Curves
Example
Example
Curves shift up & right
with increasing quality
Quality of images increases
as N & k i increase category
Similar shape in each
1/12/2009
Example
Sometimes, for fixed N,
the quality improved by decreasing k (increased contrast) image detail
Example
More vertical for higher Images with a large amount
of detail require few graygraylevels
Aliasing & Moir Patterns Moir
Aliasing & Moir Patterns Moir
Spatial Resolution Test Pattern
Voltage time
Summary
Definitions of Sampling & Quantisation Representing Digital Images Spatial & Gray-level resolution GrayPerceived Image Quality & Resolution Aliasing & Moire patterns
