Motivation towards Compression

• Digital images require huge amounts of space for

storage and large bandwidths for transmission.
– A 640 x 480 color image requires close to 1MB of
space.
• Calculate the space required for a SD ( Standard
Definition 720 * 480) movie of 2 hours running at
30 fps…!! (Answer = 224 GB)
• Imagine how movies came in two CD’s / DVD’s of
very less size than required 224 GB size…?

4/9/2020 Prof. K.P. Nayak 2

 Motivation Continued…
• Have you checked the size of any image when
its clicked by your camera and compared it to
the ones you share on social media…?
• Imagine the network congestion if all images
shared over Whatsapp or any social media did
not have any provision for compression.
• What would have been the cost of data
transmission then?

4/9/2020 Prof. K.P. Nayak 3

 Data Compression
• It aims at reducing the amount of data
required to represent a given quantity of
information.
• Data is the means by which information is
conveyed.
• Information are said to contain redundant
data.

4/9/2020 Prof. K.P. Nayak 4

 Data vs Information
• The same amount of information can be represented by
various amount of data.

• For example:-
• Your wife, Helen, will meet you at Logan Airport in Boston
at 5 minutes past 6:00 pm tomorrow night

• Your wife will meet you at Logan Airport at 5 minutes past

6:00 pm tomorrow night

• Helen will meet you at Logan at 6:00 pm tomorrow night

• All 3 statements represent the same information with
different levels of data redundancy, the first line
containing maximum redundant data.
4/9/2020 Prof. K.P. Nayak 5
 Compression Ratio

compression

Compression Ratio:

4/9/2020 Prof. K.P. Nayak 6

 Data Redundancy

• Relative data redundancy:

Example:

4/9/2020 Prof. K.P. Nayak 7

 Types of Data Redundancy

(1) Coding
(2) Interpixel
(3) Psychovisual

• Compression attempts to reduce one or

more of these redundancy types.

4/9/2020 Prof. K.P. Nayak 8

 Image Compression
• It is the art and science of reducing the
amount of data required to represent an
image.

Fig 1: Idea of Compression

4/9/2020 Prof. K.P. Nayak 9

 Coding Redundancy
• A code is a system of symbols used to represent a
body of information or set of events.
• Each piece of information or event is assigned a
sequence of code symbols called code word.
• The number of symbols in each codeword is
called its length.
• For eg:- Each pixel in a gray scale image is
represented by 8-bits binary. ( Pixel value = 255,
code = 11111111, code length per pixel= 8)
4/9/2020 Prof. K.P. Nayak 10
 Spatial & Temporal Redundancy
• Spatial Redundancy means the neighbouring
pixels are correlated in a 2-D image and hence
it is redundant to store all the pixel values.

• Temporal Redundancy is observed in case of

videos where the adjacent frames are similar
to each other i.e. the frames are correlated in
time.

4/9/2020 Prof. K.P. Nayak 11

 Psychovisual Redundancy
• The human eye does not respond with equal
sensitivity to all visual information.

• It is more sensitive to the lower frequencies

than to the higher frequencies in the visual
spectrum.

• Idea: Discard data that is perceptually

insignificant to human eyes.
4/9/2020 Prof. K.P. Nayak 12
 Types of Image Compression
• Based on the loss incurred in the compression
scheme, it is classified into the following:-
• Lossless
– Information preserving
– Low compression ratios

• Lossy
– Not information preserving
– High compression ratios

• Trade-off: image quality vs compression ratio

4/9/2020 Prof. K.P. Nayak 13

 Fidelity Criteria

• How close is to ?

• Criteria
– Subjective: based on human observers
– Objective: mathematically defined criteria

4/9/2020 Prof. K.P. Nayak 14

 Subjective Fidelity Criteria

4/9/2020 Prof. K.P. Nayak 15

 Objective Fidelity Criteria
• Root mean square error (RMS)

• Mean-square signal-to-noise ratio (SNR)

4/9/2020 Prof. K.P. Nayak 16

 Lossless Compression

4/9/2020 Prof. K.P. Nayak 17

 Lossless Methods: Taxonomy
Types of Lossless Coding Techniques
(Entropy Coding)
1.Repetitive Sequence Encoding (Ex:- Run
Length Encoding)
2.Statistical Encoding (Ex:- Huffman,
Arithmetic, LZW)
3.Lossless Predictive Coding (Ex:- DPCM)
4.Bit plane Encoding

4/9/2020 Prof. K.P. Nayak 18

 Huffman Coding (coding redundancy)

• A variable-length coding technique.

• Optimal code (i.e., minimizes the number
of code symbols per source symbol).
• Assumption: symbols are encoded one at a
time!
4/9/2020 Prof. K.P. Nayak 19
 Huffman Coding (cont’d)
• Forward Pass
1. Sort probabilities per symbol
2. Combine the lowest two probabilities
3. Repeat Step2 until only two probabilities remain.

4/9/2020 Prof. K.P. Nayak 20

 Huffman Coding (cont’d)
Backward Pass
Assign code symbols going backwards

4/9/2020 Prof. K.P. Nayak 21

 Huffman Coding (cont’d)
• Lavg (Bits per symbol) using Huffman coding:

• Lavg assuming binary codes:

4/9/2020 Prof. K.P. Nayak 22

 Huffman Coding/Decoding
• After the code has been created,
coding/decoding can be implemented using
a look-up table.
• Note that decoding is done unambiguously
in this case.

4/9/2020 Prof. K.P. Nayak 23

 Arithmetic (or Range) Coding
(coding redundancy)
• No assumption on encoding source symbols
one at a time.
– Sequences of source symbols are encoded
together.
– There is no one-to-one correspondence
between source symbols and code words.

• Slower than Huffman coding but typically

achieves better compression.
4/9/2020 Prof. K.P. Nayak 24
 Arithmetic Coding (cont’d)
• A sequence of source symbols is assigned a
single arithmetic code word which
corresponds to a sub-interval in [0,1].

• As the number of symbols in the message

increases, the interval used to represent it
becomes smaller.

• Smaller intervals require more information

units (i.e., bits) to be represented.
4/9/2020 Prof. K.P. Nayak 25
 Arithmetic Coding (cont’d)

Encode message: a1 a2 a3 a3 a4

1) Assume message occupies [0, 1)

0 1

2) Subdivide [0, 1) based on the probability of αi

3) Update interval by processing source symbols

4/9/2020 Prof. K.P. Nayak 26
 Example of Arithmetic Coding

Encode
a1 a2 a3 a3 a4

[0.06752, 0.0688)
or,
0.068

4/9/2020 Prof. K.P. Nayak 27

 Example contd…
• The message a1 a2 a3 a3 a4 is encoded using
3 decimal digits or 3/5 = 0.6 decimal digits
per source symbol.

• The entropy of this message is:

-(3 x 0.2log10(0.2)+0.4log10(0.4))=0.5786 digits/symbol

Note: finite precision arithmetic might

cause problems due to truncations!
4/9/2020 Prof. K.P. Nayak 28
 Arithmetic Decoding
1.0 0.8 0.72 0.592 0.5728
a4
0.8 0.72 0.688 0.5856 0.57152

Decode 0.572
a3

0.4 0.56 0.624 0.5728 056896

a2
a 3 a3 a1 a2 a4
0.2 0.48 0.592 0.5664 0.56768
a1
0.0 0.4
0.56 0.56 0.5664

4/9/2020 Prof. K.P. Nayak 29

 LZW Coding (Interpixel
Redundancy)
• Requires no prior knowledge of pixel
probability distribution values.

• Assigns fixed length code words to variable

length sequences.

• It is a type of Dictionary based Coding.

• Included in GIF and TIFF and PDF file formats

4/9/2020 Prof. K.P. Nayak 30
 LZW Coding
• A codebook (or dictionary) needs to be
constructed.

• Initially, the first 256 entries of the dictionary are assigned

to the gray levels 0,1,2,..,255 (i.e., assuming 8 bits/pixel)
Initial Dictionary
Dictionary Location Entry
Consider a 4x4, 8 bit image
39 39 126 126 0 0
39 39 126 126 1 1
39 39 126 126 . .
39 39 126 126 255 255
256 -

511 -
4/9/2020 Prof. K.P. Nayak 31
 LZW Coding (cont’d)
As the encoder examines image pixels,
39 39 126 126 gray level sequences (i.e., blocks) that are
not in the dictionary are assigned to a
39 39 126 126 new entry.
39 39 126 126
39 39 126 126
Dictionary Location Entry

0 0 - Is 39 in the dictionary……..Yes
1 1 - What about 39-39………….No
. .
- Then add 39-39 in entry 256
255 255
256 -
39-39

511 -

4/9/2020 Prof. K.P. Nayak 32

 Example
39 39 126 126 Concatenated Sequence: CS = CR + P
39 39 126 126
39 39 126 126 (CR) (P)
39 39 126 126

CR = empty

If CS is found:
(1) No Output
(2) CR=CS
else:
(1) Output D(CR)
(2) Add CS to D
(3) CR=P
4/9/2020 Prof. K.P. Nayak 33
 Decoding LZW

• The dictionary which was used for encoding need not

be sent with the image.

• Can be built on the “fly” by the decoder as it reads

the received code words.

4/9/2020 Prof. K.P. Nayak 34

 Run-length coding (RLC)
(Interpixel redundancy)
• Used to reduce the size of a repeating string of characters
(i.e., runs):

1 1 1 1 1 0 0 0 0 0 0 1 → (1,5) (0, 6) (1, 1)

a a a b b b b b b c c → (a,3) (b, 6) (c, 2)

• Encodes a run of symbols into two bytes: (symbol, count)

• Can compress any type of data but cannot achieve high
compression ratios compared to other compression
methods.

4/9/2020 Prof. K.P. Nayak 35

 Bit-plane coding (Interpixel redundancy)

• An effective technique to reduce inter pixel

redundancy is to process each bit plane
individually.
(1) Decompose an image into a series of binary
images.

(2) Compress each binary image (e.g., using

run-length coding)
4/9/2020 Prof. K.P. Nayak 36
 Combining Huffman Coding
with Run-length Coding

• Assuming that a message has been

encoded using Huffman coding, additional
compression can be achieved using run-
length coding.

e.g., (0,1)(1,1)(0,1)(1,0)(0,2)(1,4)(0,2)

4/9/2020 Prof. K.P. Nayak 37

 Transform based Coding
• Transform the image into a domain where
compression can be performed more
efficiently (i.e. reduce interpixel
redundancies).
• Transform types:- DFT, DCT, K-L Transform
~ (N/n)2 subimages

4/9/2020 Prof. K.P. Nayak 38

 Transform selection
-Consider an image f (x,y) of size N X N

-The forward discrete transform is defined as

N −1 N −1
T (u , v ) = 
X = 0 y =O
f ( x, y ) g ( x, y , u , v )

u , v = 0,1,......., N − 1
- The inverse discrete transform is defined as

f ( x, y ) = u =0 v =0 T (u , v)h( x, y, u , v)
N −1 N −1

x, y = 0,1,............, N − 1
4/9/2020 Prof. K.P. Nayak 39
 Example : Discrete Fourier Transform

- Forward transform kernel

2  ( ux + vy )
j
g ( x, y , u , v ) = e N

- Inverse transform kernel

ux +vy
2 ( )
1
h ( x, y , u , v ) = 2 e j N

4/9/2020 N
Prof. K.P. Nayak 40
 Discrete cosine transform(DCT)
The discrete cosine transform (DCT) gets its name from the fact
that the rows of the N x N transform matrix C are obtained as a
function of cosines.

g ( x, y , u , v ) = h ( x, y , u , v )
 (2 x + 1)u    (2 y + 1)v  
=  (u ) (v) cos   cos  
 2N   2N 
Where
1
 (u ) = for u = 0
N
2
= for u =1, 2 .........., N − 1
4/9/2020 N Prof. K.P. Nayak 41
 Why DCT

-Blocking artifacts are less pronounced in the DCT than

the DFT.

-The DCT is a good approximation of the Karhunen-Loeve

Transform(KLT) which is optimal in terms of energy
compaction.

-However unlike KLT , the DCT has image independent

basis functions.

-The DCT is used in JPEG compression standard.

4/9/2020 Prof. K.P. Nayak 42

 JPEG ( Joint Photographic Experts Group )

-> It is a compression standard for still images

-> It defines three different coding systems :

1. A lossy baseline coding system based on DCT (adequate
for most compression applications )

2. An extended system for greater compression, higher

precision, or progressive reconstruction applications

3. A lossless independent coding system for reversible

compression.

4/9/2020 Prof. K.P. Nayak 43

 Steps of JPEG Compression

4/9/2020 Prof. K.P. Nayak 44

 JPEG – Baseline Coding System

1. Computing the DCT: the image is divided into 8 X 8 blocks

The blocks of component are then transformed into
frequency components.
2. Each block is represented by 64 frequency components,
one of which (the DC component) is the average color of
the 64 pixels in the block.
3. Applying FDCT to change the blocks from spatial domain
into frequency domain.
4. Roughly speaking , F(0,0) is the average value within the
block (called DC component).
5. Other values stores the variation within a block (AC
components).
4/9/2020 Prof. K.P. Nayak 45
 Example Fast DCT
183 160 94 153 194 163 132 165
183 153 116 176 187 166 130 169
179 168 171 182 179 170 131 167
f=
177 177 179 177 179 165 131 167
178 178 179 176 179 164 130 171
179 180 180 179 182 164 130 171
179 179 180 182 183 170 129 173
180 179 181 179 181 170 130 69
4/9/2020 Prof. K.P. Nayak 46
 Example : Level Shifting
55 32 -34 25 66 35 4 37
55 25 -12 48 59 38 2 41
51 40 43 54 51 42 3 39
fs=f-128 49 49 51 49 51 37 3 39
50 50 51 48 54 36 2 43
51 52 52 51 55 36 2 43
51 51 52 54 55 42 1 45
52 51 53 51 53 42 2 41
4/9/2020 Prof. K.P. Nayak 47
 Example : Computing the DCT
312 56 -27 17 79 -60 26 -26
-38 -28 13 45 31 -1 -24 -10
-20 -18 10 33 21 -6 -16 -9
dcts=round(dct2(fs))
-11 -7 9 15 10 -11 -13 1
-6 1 6 5 -4 -7 -5 5
3 3 0 -2 -7 -4 1 2
3 5 0 -4 -8 -1 2 4
3 1 -1 -2 -3 -1 4 1

4/9/2020 Prof. K.P. Nayak 48

 Example : The Normalization Matrix

16 11 10 16 24 40 51 61

12 12 14 19 26 58 60 55

14 13 16 24 40 57 69 56
qmat=
14 17 22 29 51 87 80 62

18 22 37 56 68 109 103 77

24 35 55 64 81 104 113 92

49 64 78 87 103 121 120 101

72 92 95 98 112 100 103 99

4/9/2020 Prof. K.P. Nayak 49
 Example : Normalization Table

16 11 10 16 24 40 51 61

12 12 14 19 26 58 60 55

14 13 16 24 40 57 69 56

14 17 22 29 51 87 80 62

18 22 37 56 68 109 103 77

24 35 55 64 81 104 113 92
312/16
49 64 78 87 103 121 120 101
= 20
72 92 95 98 112 100 103 99

312 56 -27 17 79 -60 26 -26 20 5 -3 1 3 -2 1 0

-38 -28 13 45 31 -1 -24 -10 -3 -2 1 2 1 0 0 0

-1 -1 1 1 1 0 0 0
-20 -18 10 33 21 -6 -16 -9
-1 0 0 1 0 0 0 0
-11 -7 9 15 10 -11 -13 1
0 0 0 0 0 0 0 0
-6 1 6 5 -4 -7 -5 5
0 0 0 0 0 0 0 0
3 3 0 -2 -7 -4 1 2
0 0 0 0 0 0 0 0
3 5 0 -4 -8 -1 2 4 0 0 0 0 0 0 0 0
4/9/2020 Prof. K.P. Nayak 50
3 1 -1 -2 -3 -1 4 1 Quantized value
 Example : Thresholding

20 5 -3 1 3 -2 1 0

-3 -2 1 2 1 0 0 0

-1 -1 1 1 1 0 0 0

t=round(dcts. /qmat)= -1 0 0 1 0 0 0 0

0 0 0 0 0 0 0 0

0 0 0 0 0 0 0 0

0 0 0 0 0 0 0 0

0 0 0 0 0 0 0 0

4/9/2020 Prof. K.P. Nayak 51

 Zig- Zag Scanning of the Coefficients

20 5 -3 1 3 -2 1 0

-3 -2 1 2 1 0 0 0

-1 -1 1 1 1 0 0 0
t= round(dcts./qmat)= -1 0 0 1 0 0 0 0

0 0 0 0 0 0 0 0

0 0 0 0 0 0 0 0

0 0 0 0 0 0 0 0

0 0 0 0 0 0 0 0

[ 20,5,-3,-1,-2,-3,1,1,-1,-1,0,0,1,2,3,-2,1,1,0,0,0,0,0,0,1,1,0,1,0,EOB ]
4/9/2020 Prof. K.P. Nayak 52
 Decoding the Coefficients

320 55 -30 16 72 -80 51 0

-36 -24 14 38 26 0 0 0

-14 -13 16 24 40 0 0 0

t X qmat= -14 0 0 29 0 0 0 0

0 0 0 0 0 0 0 0

0 0 0 0 0 0 0 0

0 0 0 0 0 0 0 0

0 0 0 0 0 0 0 0

4/9/2020 Prof. K.P. Nayak 53

 Computing the IDCT
67 12 -9 20 69 43 -8 42

58 25 15 30 65 40 -4 47

46 41 44 40 59 38 0 49

41 52 59 43 57 42 3 42
fs_hat=round(idct2(ds_hat))

44 54 58 40 58 47 3 33

49 52 53 40 61 47 1 33

53 50 53 46 63 41 0 45

55 50 56 53 64 34 -1 57

4/9/2020 Prof. K.P. Nayak 54

 Shifting Back the Coefficients
195 140 119 148 197 171 120 170
186 153 143 158 193 168 124 175
174 169 172 168 187 166 128 177
169 180 187 171 185 170 131 170
f_hat=fs_hat+128 172 182 186 168 186 175 131 161
177 180 181 168 189 175 129 161
181 178 181 174 191 169 128 173
183 178 184 181 192 162 127 185

183 160 94 153 194 163 132 165

183 153 116 176 187 166 130 169
179 168 171 182 179 170 131 167
f= 177 177 179 177 179 165 131 167
178 178 179 176 182 164 130 171
179 180 180 179 183 164 130 171
179 179 180 182 183 170 129 173
4/9/2020 180 179 181 179 Prof.181 170
K.P. Nayak 130 169 55
 JPEG as Block Transform Encoding

4/9/2020 Prof. K.P. Nayak 56

 What is Wavelet Transform ?
-Wavelets are functions defined over a finite interval and
having an average value of zero.

- The basic idea of the wavelet transform is to represent any

arbitrary function (t) as a superposition of a set of such
wavelets or basis functions.

- These basis function or baby wavelets are obtained from a

single prototype wavelet called mother wavelet , by
contractions (scaling) and translations (shifts).

- The Discrete Wavelet Transform of a finite length signal x(n)

having N components , for example , is expressed by an N x N
matrix.
4/9/2020 Prof. K.P. Nayak 57
 Why Wavelet-based Compression?
• No need to divide input image into blocks and its basis functions
have variable length to avoid blocking artifacts.
• More robust under transmission and decoding errors.
• Better matched to the human visual system(HVS) characteristics
❖ Good frequency resolution at lower frequencies, good time
resolution at higher frequencies – good for natural images.

4/9/2020 Prof. K.P. Nayak 58

 Subband coding

-The fundamental concept behind Subband Coding (SBC) is to

split up the frequency band of a signal (image in our case) and
then to code each subband using a coder and bit rate accurately
matched to the statistics of the band.

- SBC has been used extensively first in speech coding and later
in image coding.

4/9/2020 Prof. K.P. Nayak 59

 Wavelets as filter

Essentially a multiresolution filtering process:

− HPF (Mother wavelet)

− LPF (Father scaling)

− Apply in different resolutions

4/9/2020 Prof. K.P. Nayak 60

 One Stage Filtering

Approximations and details:

- The low-frequency content is the most important part in

many applications, and gives the signal its identity.
This part is called “Approximations”.

- The high-frequency gives the ‘flavor’, and is called “Details”.

4/9/2020 Prof. K.P. Nayak 61

 Approximations and Details:

- Approximations: low- frequency components of the signal

- Details: high- frequency components

4/9/2020 Prof. K.P. Nayak 62

 Multi-level Decomposition

Iterating the decomposition process, breaks the input signal

into many lower-resolution components.
Wavelet decomposition tree is given below:-

4/9/2020 Prof. K.P. Nayak 63

 Wavelet Image Decomposition

4/9/2020 Prof. K.P. Nayak 64

 Wavelets Image decomposition
• Multilevel Decomposition
Every
512 X 512 Every row
column

256 X 256

128 X 128

4/9/2020 Prof. K.P. Nayak 65

 Decomposition and Compression

4/9/2020 Prof. K.P. Nayak 66

 Quantization
Two important observations:

1. Natural images in general have a low pass spectrum, so the

wavelet coefficients will, on average, be smaller in the higher
subbands than in the lower subbands.

2. Large wavelet coefficients are more important than smaller

wavelet coefficients.

4/9/2020 Prof. K.P. Nayak 67

 Example

• Here is a two-stage wavelet decomposition of an image.

Notice the large number of zeros (black):

Matlab toolbox
4/9/2020 Prof. K.P. Nayak 68
 Comparison

JPEG (DCT based) JPEG-2000 (Wavelet

4/9/2020 Prof. K.P. Nayak 69
based)
 Comparison JPEG-2000(1.83KB)

Original(979 KB)

JPEG (6.21 KB)

4/9/2020 Prof. K.P. Nayak 70
 References
• Gonzalez R.C. and Woods R.E, “Digital Image
Processing”, Pearson
• https://www.youtube.com/playlist?list=PLuv3
GM6-gsE08DuaC6pFUvFaDZ7EnWGX8 (NPTEL
video tutorial Part-1 by Dr. P.K. Biswas, IIT
Kharagpur)
• https://www.youtube.com/playlist?list=PL1F0
76D1A98071E24 (NPTEL video tutorial Part-2
by Dr. P.K. Biswas, IIT Kharagpur)

4/9/2020 Prof. K.P. Nayak 71