Chapter3 DIP English Nou

Digital image processing
quantization
Chapter 3. Image sampling and
IMAGE SAMPLING AND IMAGE QUANTIZATION

1. Introduction
2. Sampling in the two-dimensional space
Basics on image sampling
The concept of spatial frequencies
Images of limited bandwidth
Two-dimensional sampling
Image reconstruction from its samples
The Nyquist rate. The alias effect and spectral replicas superposition
The sampling theorem in the two-dimensional case
Non-rectangular sampling grids and interlaced sampling
The optimal sampling
Practical limitations in sampling and reconstruction
3. Image quantization
4. The optimal quantizer
The uniform quantizer
5. Visual quantization
Contrast quantization
Pseudo-random noise quantization
Halftone image generation
Color image quantization

quantization
1.
Introduction
Fig 1 Image sampling and quantization / Analog image display

quantization
Sampling in the two-dimensional space

Basics on image sampling
f(x,y)
2.

quantization
The concept of spatial frequencies
- Grey scale images can be seen as a 2-D generalization of time-varying signals (both in the analog and
in the digital case); the following equivalence applies:
1-D signal (time varying)
2-D signal (grey scale image)
Time coordinate t
Space coordinates x,y
Instantaneous value: f(t)
Brightness level, point-wise: f(x,y)
A 1-D signal that doesnt vary in time (is constant)

= has 0 A.C. component, and only a D.C.
component
A perfectly uniform image (it has the same

brightness in all spatial locations); the D.C.
component = the brightness in any point
The frequency content of a 1-D signal is

proportional to the speed of variation of its
instantaneous value in time:
max ~ max(df/dt)
The frequency content of an image (2-D signal) is

proportional to the speed of variation of its
instantaneous value in space:
max,x ~ max(df/dx); max,y ~ max(df/dy)
=> max,x , max,y = spatial frequencies

Discrete 1-D signal: described by its samples => a
vector: u=[u(0) u(1) u(N-1)], N samples; the
position of the sample = the discrete time moment
Discrete image (2-D signal): described by its

samples, but in 2-D => a matrix: U[MN],
U={u(m,n)}, m=0,1,,M-1; n=0,1,,N-1.
The spectrum of the time varying signal = the real

part of the Fourier transform of the signal, F();
=2.
The spectrum of the image = real part of the

Fourier transform of the image = 2-D
generalization of 1-D Fourier transform, F(x,y)
x=2x; y=2y

quantization
Images of limited bandwidth

Limited bandwidth image = 2-D signal with finite spectral support:
F(x, y) = the Fourier transform of the image:
F x , y f x, y e

j x x j y y
dxdy f x, y e
j 2 x x y y
dxdy.
F ( x , y ) 0,| x | x 0 ,| y | y 0
Thespectrumofalimitedbandwidthimageanditsspectralsupport

quantization
Two-dimensional sampling (1)

The common sampling grid = the uniformly spaced, rectangular grid:
g ( x, y ) ( x, y )
( x mx, y ny )
m n
Image sampling = read from the original, spatially continuous, brightness function f(x,y), only in the black
dots positions ( only where the grid allows):
f ( x, y ), x mx, y ny
,
0
,
otherwise
f s ( x, y )
m, n Z.
f s ( x, y ) f ( x, y ) g ( x, y ) ( x, y )
f (mx, ny ) ( x mx, y ny )
m n

quantization

Question: How to choose the values x, y to achieve:
-the representation of the digital image by the min. number of samples,
-at (ideally) no loss of information?
(I. e.: for a perfectly uniform image, only 1 sample is enough to completely represent the image =>
sampling can be done with very large steps; on the opposite if the brightness varies very sharply => very
many samples needed)
The sampling intervals x, y needed to have no loss of information depend on the spatial frequency
content of the image.
Sampling conditions for no information loss derived by examining the spectrum of the image by
performing the Fourier analysis:
f s ( x, y ) f ( x, y ) g ( x, y ) ( x, y ) Fourier transform :

FS ( x , y ) f S ( x, y ) e j 2 x x e
j 2 y y
dxdy f ( x, y ) g ( x, y ) ( x, y ) e j 2 x x e
j 2 y y
dxdy.
The sampling grid function g(x, y) is periodical with period (x, ly) => can be expressed by its Fourier
k
j 2 x j 2 y
series expansion:
y
x
g ( x, y ) ( x, y )
a(k , l ) e
k l
where :
k
j 2 x j 2 y
1 1 x y
y
x e
a(k , l )
g ( x, y ) ( x, y )e
dxdy.
x y 0 0

quantization

Since:
1, if x y 0
,
0
,
otherwise
for ( x, y ) [0; x) [0; y ), g ( x, y ) ( x, y )

a (k , l )
1 1
, k , l ; .
x y
Therefore the Fourier transform of fS is:
2 k x
2 l y
j
j
1
x
FS ( x , y ) f ( x, y )
e
e y

k l x y
j 2 x x
e
e
j 2 y y
l
k
j 2 y y
j 2 x x
1
y
x e
FS ( x , y )
f ( x, y ) e
x y k l
dxdy
l
k
j 2 y y
j 2 x x
1
y
x e
FS ( x , y )
f ( x, y ) e
dxdy
x y k l
1
k
l
FS ( x , y )
F x , y .
x y k l
x
y
dxdy
The spectrum of the sampled image = the collection of an infinite number of scaled spectral replicas

quantization
Original image
Original image spectrum 3D
Original image spectrum 2D
2-D rectangular sampling grid

Sampled image spectrum 3D
Sampled image spectrum 2D

quantization
Image reconstruction from its samples
xs 2 x 0 , ys 2 y 0
H ( x , y )
1
1
,y
2 x 0
2 y 0
1
,( x , y )
( xs ys )
0,
otherwise
~
F ( x , y ) H ( x , y ) Fs ( x , y ) F ( x , y )
~
f x, y h x, y f s x, y
~
f x, y
f s mx, ny h x mx, y ny
m n
Let us assume that the filtering region R is rectangular, at the middle distance between two spectral
replicas:
ys
xs
1
sin x xs sin y ys
,
and
x
y
H ( x , y ) ( xs ys )
2
2 h x, y
y ys
xs
0,
otherwise
~
f x, y
sin x xs m sin y ys n
f
m
x
,
n
y
h
x
x
,
y
f
m
x
,
n
y ys n
xs
m n
m n

quantization
~
f x, y
f s mx, ny sinc x xs m sinc y ys n , where sinc a
m n
sin a
a
Since the sinc function has infinite extent => it is impossible to implement in practice the ideal LPF
it is impossible to reconstruct in practice an image from its samples without error if we sample it
at the Nyquist rates.
Practical solution: sample the image at higher spatial frequencies + implement a real LPF (as close to
the ideal as possible).
1-D sinc function

2-D sinc function

quantization
The Nyquist rate. The aliasing. The fold-over frequencies
xs 2 x 0 , ys 2 y 0
The Moire effect
Note: Aliasing may also appear in the reconstruction process,

due to the imperfections of the filter!
How to avoid aliasing if cannot increase the sampling frequencies?
By a LPF on the image applied prior to sampling!
Jagged boundaries

quantization
Non-rectangular sampling grids. Interlaced sampling grids

y
1/2
F(x, y)=1
1/ 2
x
1/2
-1/2
-1/2
a) Image spectrum
x
-1
d) The spectrum using G1
e) The spectrum using G2
Interlacedsampling
Optimal sampling = Karhunen-Loeve expansion:
f ( x, y)
m ,n 0
m ,n
m ,n

quantization
Image reconstruction from its samples in the real case
The question is: what to fill in the interpolated (new) dots?

Several interpolation methods are available; ideally sinc function in the spatial domain; in
practice simpler interpolation methods (i.e. approximations of LPFs).

Image interpolation filters:

quantization
Image interpolation examples:
quantization
1. Rectangular (zero-order) filter, or nearest neighbour filter, or box filter:

1/x
-x/2
Original
x/2
Sampled
Reconstructed
quantization
2. Triangular (first-order) filter, or bilinear filter, or tent filter:

1/x
-x
Original
Sampled
Reconstructed
quantization
3. Cubic interpolation filter, or bicubic filter begins to better approximate the sinc function:
2x
Original
Sampled
Reconstructed

quantization
Practical limitations in image sampling and reconstruction
Fig. 7 The block diagram of a real sampler & reconstruction (display) system
Fig. 8 The real effect of the interpolation

quantization
3. Image quantization
3.1. Overview
Fig. 9 The quantizers transfer function

quantization
3.2. The uniform quantizer

The quantizers design:
Denote the input brightness range:
u lmin ; LMax
Let B the number of bits of the quantizer => L=2B reconstruction levels
E.g. B=2 => L=4
The expressions of the decision levels:
Uniform quantizer transfer function
t1 l min ; t L 1 LMax
L
l
t
t
q L 1 1 ,t k t k 1 q q Max min
L
L
Reconstruction levels
t k t k 1 t k 1 t k constant q
r4=224
r3=160
r2=96
The expressions of the reconstruction levels:

r1=32
t t
q
rk k k 1 rk t k
2
2
t1=0
t2=64
t3=128
Decision levels
t4=192
Computation of the quantization error: for a given image of size MN pixels,

U non-quantized, and U quantized => we estimate the MSE:
t5=256

quantization
Examples of uniform quantization and the resulting errors:

B=1 => L=2
Non-quantized image
Quantized image
r2=192
r1=64
t1=0
t2=128
Decision levels
t3=256
Quantization error; MSE=36.2
1000
900
800
700
600
500
400
300
200
100
0
0
50
100
150
200
250
The histogram of the non-quantized image

quantization

B=2 => L=4
Non-quantized image
Quantized image
r4=224
r3=160
r2=96
r1=32
t1=0
t2=64
t3=128
Decision levels
t4=192
t5=256
Quantization error; MSE=15
1000
900
800
700
600
500
400
300
200
100
0
0
50
100
150
200
250

quantization

B=3 => L=8; false contours present
Non-quantized image
Quantized image
r8=240
r7=208
r6=176
r5=144
r4=112
r3=80
r2=48
r1=16
t1=0
t2=32
t3=64
t4=96 t5=128 t6=160

Decision levels
t7=192
t8=224
t9=256
Quantization error; MSE=7.33
1000
900
800
700
600
500
400
300
200
100
0
0
50
100
150
200
250

quantization
3.2. The optimal (MSE) quantizer (the Lloyd-Max quantizer)
(tk rk 1 ) 2 (t k rk ) 2 hu (t k ) 0
t k
t k 1
2 (u rk )hu (u )du 0 1 k L
rk
t
k
t k 1
r r
t k k k 1
2
rk
uhu (u)du
tk
t k 1
E u | u k
hu (u)du
tk
pu (u) pu (tj ),
tj
z k t1
A
t k 1
1
(t j t j 1 ),
2
[hu (u )]
1 / 3
du
t1
t L 1
[hu (u )]
t1
1 / 3
t1
du
t j u t j 1

quantization
12 L2
t L1
t1
[hu (u )]
1/ 3
du
pu(u)
u
t1
t2
(u ) 2
hu (u )
exp
2
2
2
2
(Laplacian)
1
2
(
tj
tj+1
tL+1
h
(
u
)
exp u
u
(Gaussian)
,2 or
variance , - mean)

quantization
Examples of optimal quantization and the quantization error:

B=1 => L=2
Non-quantized image
Nivelele de reconstructie
Functia de transfer a cuantizorului optimal
Quantized image
r2=153
r1=24
t1=0
t2=89
Nivelele de decizie
t3=256
The quantization error;

MSE=19.5
1000
900
The evolution of MSE

in the optimization, starting
from the uniform quantizer
800
38
700
36
600
34
500
32
400
30
300
28
200
26
100
24
22
250
20
The non-quantized image histogram
18
50
100
150
200

quantization

B=2 => L=4
Non-quantized image
Quantized image
r4=181
r3=156
r2=115
r1=20
t1=0
t2=68
t3=136 t4=169
Nivelele de decizie
t5=256
The quantization error;

MSE=9.6
1000
The evolution of MSE

900
800
15
700
600
14
500
13
400
300
12
200
11
100
0
10
50
100
150
200
250
10

quantization

B=3 => L=8
Non-quantized image
Quantized image
r8=224
r7=181
r6=165
r5=147
r4=125
r3=101
r2=54
r1=14
t1=0
t2=34
t3=78
t4=113 t5=136 t6=156t7=173

Nivelele de decizie
t8=203
t9=256
The quantization error; The evolution of MSE

MSE=5
1000
900
800
700
7.5
600
7
500
400
6.5
300
6
200
100
5.5
0
5
50
100
150
200
250
4.5
10
12
14

quantization
3.3. The uniform quantizer = the optimal quantizer for the uniform grey level
distribution:
1
, t1 u t L 1
hu (u ) t L 1 t1
0 otherwise
(t k21 t k2 ) t k 1 t k
rk
2(t k 1 t k )
2
t k 1 t k 1
t k t k 1 t k 1 t k constant q
2
t t
q
q L 1 1 ,
t k t k 1 q ,
rk t k
L
2
tk
q /2
1
q2
2
u du
q q/ 2
12
2 2B , therefore SNR 10log10 2 2B 6 B dB

2
u

quantization
3.4. Visual quantization methods

In general if B<6 (uniform quantization) or B<5 (optimal quantization) => the "contouring"
effect (i.e. false contours) appears in the quantized image.

The false contours (contouring) = groups of neighbor pixels quantized to the same value <=>
regions of constant gray levels; the boundaries of these regions are the false contours.
The false contours do not contribute significantly to the MSE, but are very disturbing for the
human eye => it is important to reduce the visibility of the quantization error, not only the MSQE.
Solutions: visual quantization schemes, to hold quantization error below the level of visibility.
Two main schemes: (a) contrast quantization; (b) pseudo-random noise quantization
Uniform quantization, B=4
Optimal quantization, B=4

quantization
3.4. Visual quantization methods

a. Contrast quantization
The visual perception of the luminance is non-linear, but the visual perception of contrast is linear
uniform quantization of the contrast is better than uniform quantization of the brightness
contrast = ratio between the lightest and the darkest brightness in the spatial region
just noticeable changes in contrast: 2% => 50 quantization levels needed 6 bits needed with a
uniform quantizer (or 4-5 bits needed with an optimal quantizer)
c ln(1 u ),0 u 1; typ. 6...18, / ln(1 )

or
c u ; typ. 1; 1 / 3

quantization
Examples of contrast quantization:

For c=u1/3:
The transfer function of the contrast quantizer

250
1
0.9
200
0.8
0.7
0.6
0.5
150
100
0.4
0.3
50
0.2
0.1
0
0
t2=3.9844
t1=0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
t3=31.875
t4=107.5781
Decision levels
1000
900
800
700
600
500
400
300
200
100
0
0
50
100
150
200
250
t5=255

quantization
Examples of contrast quantization:

The transfer function of the contrast quantizer
For the log transform:
250
5
4.5
200
4
3.5
3
2.5
2
150
100
1.5
50
1
0.5
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
1000
900
800
700
600
500
400
300
200
100
0
0.8
0.9
0
t1=0
t2=46.0273
t3=102.2733
t4=171.0068
Decision levels
t5=255

quantization
b. Pseudorandom noise quantization (dither)
Large dither amplitude
Prior to dither subtraction
Small dither amplitude

quantization

quantization
Halftone images generation
Fig.14Digitalgenerationofhalftoneimages
Demo: http://markschulze.net/halftone/index.html
Fig.15Halftonematrices

quantization
Fig.3.16

quantization
Color images quantization
Fig.17Colorimagesquantization

Chapter3 DIP English Nou

Hochgeladen von

Dokumentinformationen

Originalbeschreibung:

Originaltitel

Copyright

Verfügbare Formate

Dieses Dokument teilen

Dokument teilen oder einbetten

Freigabeoptionen

Stufen Sie dieses Dokument als nützlich ein?

Sind diese Inhalte unangemessen?

Copyright:

Verfügbare Formate

Chapter3 DIP English Nou

Hochgeladen von

Copyright:

Verfügbare Formate

Digital image processing

Chapter 3. Image sampling and

IMAGE SAMPLING AND IMAGE QUANTIZATION

Digital image processing

Chapter 3. Image sampling and

Fig 1 Image sampling and quantization / Analog image display

Digital image processing

Chapter 3. Image sampling and

Sampling in the two-dimensional space

Digital image processing

Chapter 3. Image sampling and

The concept of spatial frequencies

1-D signal (time varying)

2-D signal (grey scale image)

Space coordinates x,y

Instantaneous value: f(t)

Brightness level, point-wise: f(x,y)

A 1-D signal that doesnt vary in time (is constant)

A perfectly uniform image (it has the same

The frequency content of a 1-D signal is

The frequency content of an image (2-D signal) is

=> max,x , max,y = spatial frequencies

Discrete image (2-D signal): described by its

The spectrum of the time varying signal = the real

The spectrum of the image = real part of the

Digital image processing

Chapter 3. Image sampling and

Images of limited bandwidth

Digital image processing

Chapter 3. Image sampling and

Two-dimensional sampling (1)

Digital image processing

Chapter 3. Image sampling and

Two-dimensional sampling (2)

Digital image processing

Chapter 3. Image sampling and

Two-dimensional sampling (3)

for ( x, y ) [0; x) [0; y ), g ( x, y ) ( x, y )

Therefore the Fourier transform of fS is:

Digital image processing

Chapter 3. Image sampling and

Original image spectrum 3D

Original image spectrum 2D

2-D rectangular sampling grid

Sampled image spectrum 2D

Digital image processing

Chapter 3. Image sampling and

Image reconstruction from its samples

Digital image processing

Chapter 3. Image sampling and

f s mx, ny sinc x xs m sinc y ys n , where sinc a

1-D sinc function

Digital image processing

Chapter 3. Image sampling and

The Nyquist rate. The aliasing. The fold-over frequencies

The Moire effect

Note: Aliasing may also appear in the reconstruction process,

Digital image processing

Chapter 3. Image sampling and

Non-rectangular sampling grids. Interlaced sampling grids

d) The spectrum using G1

e) The spectrum using G2