National University of Sciences and

Technology (NUST)
School of Electrical Engineering and Computer Science
(SEECS)

Digital Image Processing

1 Khawar Khurshid
 Sampling
Quantization
Interpolation

2 Khawar Khurshid
 Sampling and Quantization
Sampling:
Digitization of the spatial coordinates (x,y)

Quantization:
Digitization in amplitude (also known as gray
level quantization)

• 8 bit quantization: 28 =256 gray levels

(0: black, 255: white)

• 1 bit quantization: 21 = 2 gray levels

(0: black, 1: white)

3 Khawar Khurshid
 Sampling

4 Khawar Khurshid
 Sampling

5 Khawar Khurshid
 Quantization – (Example)

Return change using only

these notes
6 Khawar Khurshid
 Quantization – (Example)

50
For Rs. 3 Return 5 45

Change you return

40
For Rs. 7 Return 5
35

For Rs. 9 Return 10

30
25

For Rs. 12 Return 10 20

15

For Rs. 23 Return 25 10

5
… … 0
0 5 10 15 20 25 30 35 40 45 50

Actual change

7 Khawar Khurshid
 Quantization – (Example)

Discrete color output

Continuous colors mapped to a

finite, discrete set of colors.

Continuous color input

8 Khawar Khurshid
 Quantization – (Example)

16 million colors

16 colors

9 Khawar Khurshid
 Sampling and Quantization

Pixel Grid

Real Image Sampled Quantized Sampled &

Quantized

10
Khawar Khurshid
 Sampling and Quantization

11
Khawar Khurshid
 Sampling and Quantization

12
Khawar Khurshid
 Sampling and Quantization

256x256 128x128 64x64 32x32

256 gray levels

16 gray levels 8 gray levels 4 gray levels Binary Image

13 Khawar Khurshid
 Interpolation

14 Khawar Khurshid
 Interpolation

Why?

• Changing Perspective
• Rotation
• Resizing
• Warping

15
Khawar Khurshid
 Nearest Neighbor Replacement

Simply Replicate the value from neighboring

pixels.
1 0 1

1 0 1
1 1 0 1 1 0
1 0 1

1 0 1
16
Khawar Khurshid
 Nearest Neighbor Replacement

Adds blocky effect.

1 1 0 0 0 1 1
1 1 0 0 0 1 1
1 0 1 1 1 1 1 1 0 0
1 1 0 1 1 1 1 1 0 0
1 0 1 1 1 1 1 1 0 0
1 1 0 0 0 1 1
1 1 0 0 0 1 1
17
Khawar Khurshid
 1-D Interpolation

Interpolation works by using known data to

estimate values at unknown points.

18
Khawar Khurshid
 1-D Interpolation

Linear Interpolation: Fit a linear function piecewise

between the points.

19
Khawar Khurshid
 1-D Interpolation

Quadratic: Fit a Quadratic Polynomial between the three

points.

20
Khawar Khurshid
 1-D Interpolation

Fitting Polynomial to data points.

21
Khawar Khurshid
 1-D Interpolation
Degree Polynomial How it looks Req. Pts.
Zero
f(x) = 0 None
Polynomial
horizontal line
0 f(x) = a0 with y-intercept 1
a0
an oblique line
1 (Linear) f(x) = a0 + a1x with y-intercept 2
a0 and slope a1
2 Parabola
f(x) = a0 + a1x + a2x2 3
(Quadratic)
4
f(x) = a0 + a1x + a2x2+
3 (Cubic) Spline
a3x3

• No. of unknowns in a polynomial of degree n-1 = n

• Select n nearest known neighbours to generate n equations
• Solve the linear system of equations to estimate the constants
• Simply put the unknown x value and interpolate the y value
22
Khawar Khurshid
 Linear Interpolation

• Equation of a line

y1  y0
y x  x0   y0
x1  x0

• Interpolation using Linear Polynomial

– y = a0 + a1x
– Requires two unknowns
23
Khawar Khurshid
 Bi-linear Interpolation

• Key Points:

– Q11 = (x1, y1), Q12 = (x1, y2),

– Q21 = (x2, y1), Q22 = (x2, y2).

• Linear interpolation in the x-

direction followed by linear
interpolation in y-direction

24
Khawar Khurshid
 Interpolation

Nearest Neighbor

Bi-Linear

25
Khawar Khurshid
 Interpolation

215

128

26
Khawar Khurshid
 Interpolation

27
Khawar Khurshid
 Interpolation – (Comparison)

We’ll enlarge this image

by a factor of 4 …

… via bilinear interpolation

and compare it to a nearest
neighbor enlargement.

28
Khawar Khurshid
 Interpolation – (Comparison)

Original
Image

To better see what happens, we’ll look at the parrot’s eye.

29
Khawar Khurshid
 Interpolation – (Comparison)

Pixel replication Bilinear interpolation

30
Khawar Khurshid
 Interpolation – (Comparison)

Pixel replication Bilinear interpolation

31
Khawar Khurshid
 Interpolation – (Non Integer)

Zoom in on a
section for a
closer look at
the process

Example:
resize to 3/7
of the original
32
Khawar Khurshid
 Interpolation – (Non Integer)

3/7 resize

Zoom in for a
better look

33
Khawar Khurshid
 Interpolation – (Non Integer)

3/7 resize

Outlined in
blue: 7x7
pixel squares

34
Khawar Khurshid
 Interpolation – (Non Integer)

3/7 resize

In yellow: 3 pixels
for every 7 rows,
3 pixels for every
7 cols.

35
Khawar Khurshid
 Interpolation – (Non Integer)

3/7 resize

Keep the
highlighted
pixels…

36
Khawar Khurshid
 Interpolation – (Non Integer)

3/7 resize

… don’t keep
the others.

37
Khawar Khurshid
 Interpolation – (Non Integer)

3/7 resize

Copy them into

a new image.

38
Khawar Khurshid
 Interpolation – (Non Integer)

3/7 resize

Copy them into

a new image.

39
Khawar Khurshid
 Interpolation – (Non Integer)

3/7 times the

linear dimensions
of the original

40
Khawar Khurshid
 Interpolation – (Non Integer)

Original
image

Detail of
resized image

Resize to 3/7 of
the original dims.

41
Khawar Khurshid
 Interpolation – (Non Integer)

7/3 resize

Original
image
Pixels spread out … then filled in.
for a 7/3 resize …

42
Khawar Khurshid
 Interpolation – (Non Integer)

7/3 resize

Each 3x3 block … is spread out over

of pixels from a 7x7 block here
here …

Original
image Detail

43
Khawar Khurshid
 Interpolation – (Non Integer)

7/3 resize

3x3 blocks
distributed over
7x7 blocks

44
Khawar Khurshid
 Interpolation – (Non Integer)

7/3 resize

Empty pixels filled

with color from non-
empty pixel

45
Khawar Khurshid
 Interpolation – (Non Integer)

7/3 resize

Empty pixels filled

with color from non-
empty pixel

46
Khawar Khurshid
 Interpolation – (Non Integer)

Original
image
7/3 resized

47
Khawar Khurshid
 Interpolation
• Non-adaptive algorithms

– Examples
• nearest neighbour, bilinear, bicubic, spline etc.
– Depending on their complexity, these use different number of adjacent
pixels when interpolating.
– The more adjacent pixels they include, the more accurate they can become,
but this comes at the expense of much longer processing time.

• Adaptive algorithms

– Many of these apply a different version of their algorithm (on a pixel-by-

pixel basis) when they detect the presence of an edge-- aiming to minimize
unsightly interpolation artifacts in regions where they are most apparent.

– These algorithms are primarily designed to maximize artifact-free detail in

enlarged photos, so some cannot be used to distort or rotate an image.

48
Khawar Khurshid
 End
Sampling, Quantization,
Interpolation

49 Khawar Khurshid