CSE4227 Digital Image Processing

Chapter 8 – Image Compression

Dr. Kazi A Kalpoma

Professor, Department of CSE
Ahsanullah University of Science & Technology (AUST)

Contact: kalpoma@aust.com
Google Class code: o5sjihh

CSE | AUST Fall 2022

Today’s Contents
❑ Why do we need compression?
❑ Compression Ratio and Relative Data Redundancy
❑ How can we implement compression?
❑ Coding redundancy
❑ Spatial and temporal redundancy
❑ Irrelevant information
❑ Image Entropy
❑ Lossy and Loss-free Compression
❑ Different Compression Techniques
❖ Huffman coding
❖ Golomb coding
❖ Arithmetic coding
❖ LZW coding
❖ Run length coding
Chapter 8 from R.C. Gonzalez and R.E. Woods, Digital Image Processing
(3rd Edition), Prentice Hall, 2008 [ Section 8.1, 8.2 ]
 Fundamentals
• The term Data Compression refers to the process of reducing
the amount of data required to represent a given quantity of
information

• Data and Information are not the same thing in image

• Various amount of data can be used to represent the same

information

• Data might contain elements that provide irrelevant or

repeated information : Data Redundancy

• Data redundancy is a central issue in image compression.

• It is not an abstract concept but mathematically quantifiable
entity
 Why do we need compression?

❑ Data storage

❑ Data transmission
Applications that require image compression
are many and varied such as:

1. Internet,
2. Businesses,
3. Multimedia,
4. Satellite imaging,
5. Medical imaging
 Why Images Are Compressed?

• Let a SD TV 2-hour color sequence

– each of size 720X480
– frame rate: 30 frames/sec
 Why Images Are Compressed?

• Let a SD TV 2-hour color sequence

– each of size 720X480
– frame rate: 30 frames/sec

• For a single second, the amount of data to be

accessed is
pixels bytes
30 frames (720  480) 3  31MB
frame pixel
 Why Images Are Compressed?

• For a 2-hour long SD TV video, the data is

bytes 2 sec
31,104,000  (60)  2 hours
sec hour
 2.24 1011 bytes
= 224 GB

• Twenty seven 8.5 GB dual-layer DVDs (12 cm disks) are needed

to store it.
• To put on a single DVD, must be compressed by a factor of 26.3
 Why Images Are Compressed?
• High Definition (HD) TV color sequence
– each of size 1920X1080 pixels
Compression must be higher
 Why Images Are Compressed?
• High Definition (HD) TV color sequence
– each of size 1920X1080 pixels
Compression must be higher

• Web Images
– color images of size 128 X 128 through 56Kbps to
12Mbps connections
Requires 7.0 to 0.03 seconds to download
 Why Images Are Compressed?
• High Definition (HD) TV color sequence
– each of size 1920X1080 pixels
Compression must be higher
• Web Images
– color images of size 128 X 128 through 56Kbps to
12Mbps connections
Requires 7.0 to 0.03 seconds to download
• 1 GB Flash memory and 8 Megapixel Digital
camera
Can store at most ~41 uncompressed images
 Data Redundancy
❑ Let n1 and n2 denote the number of information carrying units in two
data sets that represent the same information

❑ The relative redundancy RD is define as :

1
RD = 1 −
CR
where CR, commonly called the compression ratio, is
n
CR = 1
n2

If n1 = n2 , CR=1 and RD=0 no redundancy

If n1 >> n2 , CR → ∞ and RD → 1 high redundancy
If n1 << n2 , CR → 0 and RD → ∞ undesirable
❑ A compression ration of 10 (10:1) means that the first data set has 10 information
carrying units (say, bits) for every 1 unit in the second (compressed) data set.
 How can we implement compression?

• Coding redundancy
Most 2-D intensity arrays contain more bits than are
needed to represent the intensities

• Spatial and temporal redundancy

Pixels of most 2-D intensity arrays are correlated
spatially and video sequences are temporally correlated

• Irrelevant information
Most 2-D intensity arrays contain information that is
ignored by the human visual system
 Coding Redundancy

❑ Code is system of symbols to represent

info
❑ Codeword or sequence of symbols
represents a piece of info
❑ The number of symbols required is its
length
❑ Try to minimize codeword length
 Variable length Codeword representations
for a piece of info
 Optimal information Coding
❑ Recall from the histogram calculations
h(rk ) nk
p(rk ) = =
n n
where p(rk) is the probability of a pixel to have a certain
value rk

❑ If the number of bits used to represent rk is l(rk), then

average number of bits required to represent each pixel is
L −1
Lav =  l (rk )( p(rk )
k =0
k = L −1
Total bit requirment is MNLavg = MN  l (r ) p(r )
k =0
k k
 Coding Redundancy: Example

Total Bit
required is
Computer generated
256x256x8 bit image with
coding redundancy =256X256X8
 Coding Redundancy: Example
Total
=256X256X1.81 bits

C =8/1.81=~4.42
Computer generated
256x256x8 bit image with
coding redundancy R=1-1/C=.774

77.4 %
Lavg =.25X2+.47X1+.25X3+.03X3=1.81 bits
Coding Redundancy
Variable-Length Coding

2nd Example:

7
Lav =  l (rk )( p (rk )
k =0

= 2(019) + 2(0.25) + 3(0.16) + ... + 6(0.02)

= 2.7 bits
3
C R = = 1.11
2
1
RD = 1 − = 0.099 9.9% data redundant.
1.11
 Spatial Redundancy

• Intensities of each pixel may correlate to

its neighbors

• Many info is unnecessarily replicated

• Try to minimize unnecessary data

 Spatial and Temporal Redundancy
Example

Computer generated 256x256x8 bit Histogram

image with spatial redundancy

• Histogram is equiprobable
• Vertically, pixels are independent
• Horizontally, they are maximally correlated
• The image cannot be compressed using variable length
coding
 Spatial and Temporal Redundancy
Example
Original Bit required
=256X256X8
In Run length
=(256+256)X8
C =(256X256X8)/(256+256)X8
Computer generated 256x256x8 bit =128:1
image with spatial redundancy

• The image can be represented as a sequence of run length

pair : intensity value and num of pixels with that intensity
 Irrelevant Information

• Many images has info that are ignored by the

Human Visual System (HVS)

• Try to remove these extraneous data

 Irrelevant Information

Computer generated 256x256x8 bit

image with irrelevant information

❑ Can be represented using a single average gray level.

C =256X256X8:8
=65536:1
 Irrelevant Information

Computer generated 256x256x8 bit

image with irrelevant information

❑ However, this coarse quantization may remove useful

invisible information
Redundancies We Tried to Remove

• Coding redundancy
• Spatial redundancy
• Irrelevant information
How many bits do we really need
to represent image-information?

Information theory: does it have any answer?

 Measuring Image Information
Information Theory Review
A random event E with probability P(E)
carries I(E) units of information,

1
I ( E ) = log = − log P( E )
P( E )

• Higher the uncertainty, higher the information content

 Information theory
If we have a source of random events from a discrete
set of events {a1, a2, a3,. . ., aJ} with probabilities,
P(a1), P(a2), P(a3),. . ., P(aJ) then the average
information per event or the entropy of the source,

J
H =  − P(a j ) log P(a j )
j =1
 Calculating Image Entropy

If we consider the pixel intensities as random events,

then the intensity histogram is the approximation of
the probabilities

~ L −1
H =  − Pr (rk ) log P(rk )
k =1

✓Entropy is the measurement of the average information

in an image
 Image Entropy
• Image entropy is a quantity which is used to describe the
amount of information which must be coded for by a
compression algorithm.

• Low entropy images, such as those containing a lot of black

sky, have very little contrast and large runs of pixels with the
same or similar DN values.

• An image that is perfectly flat will have an entropy of zero.

Consequently, they can be compressed to a relatively small
size.

• On the other hand, high entropy images such as an image of

heavily cratered areas on the moon have a great deal of
contrast from one pixel to the next and consequently cannot
be compressed as much as low entropy images.
 Image Entropy: Example
~ L −1
H =  − Pr (rk ) log P(rk )
k =1

= −[0.25 log 2 0.25 + 0.47 log 2 0.47

+ 0.25 log 2 0.25 + 0.03 log 2 0.03]
 −[0.25(−2) + 0.47(−1.09)
+ 0.25(−2) + 0.03(−5.06)]
 1.6614 bits/pixel
rk Pr(rk)
 Image Entropy: Example

H=1.6614 H=8 H=1.566

✓The amount of Entropy and thus information in an image

is far from intuitive.
 Image Entropy: Example

H=1.6614 H=8 H=1.566

❖Visual perception and entropy (information) are

far from compliance
Entropy - Example
Entropy - Example
 Compression Types

Compression

Error-Free Compression
Lossy Compression
(Loss-less)
Different Error Free Compression
Techniques

❑ Variable length coding

❖ Huffman coding
❖ Golomb coding
❖ Arithmetic coding

❑ LZW coding
❑ Bit plane coding
❖ Constant area coding
❖ Run length coding
 Huffman Coding

• The most popular technique for removing coding

redundancy is due to Huffman (1952)

• Huffman Coding yields the smallest number of code

symbols per source symbol

• The resulting code is optimal

 Steps for Huffman Algorithm
• The Huffman algorithm can be described in five
steps:

1. Find the gray level probabilities for the image by

finding the histogram
2. Order the input probabilities (histogram
magnitudes) from smallest to largest
3. Combine the smallest two by addition
4. GOTO step 2, until only two probabilities are left
5. By working backward along the tree, generate
code by alternating assignment of 0 and 1
40
 Huffman Coding: Example

Consider a 3 bit image with the following probabilities:

Source Symbol Probability

0.1
0.4
0.06
0.1
0.04
0.3
 Huffman Coding: Example
Consider a 3 bit image with the following probabilities:
 Huffman Coding: Example

Consider a 3 bit image with the following probabilities:

 Huffman Coding: Example
Consider a 3 bit image with the following probabilities:

Source Reduction
 Huffman Coding: Example

Consider a 3 bit image with the following probabilities:

Source Reduction
 Huffman Coding: Example
Consider a 3 bit image with the following probabilities:
 Huffman Coding: Example

Huffman Code assignment

with 0 and 1:
 Huffman Coding

Huffman Code assignment:

 Huffman Coding

Huffman Code assignment:

 Huffman Coding

Huffman Code assignment:

 Huffman Coding

Huffman Code assignment:

Huffman Coding

0
0 1
0 1
0
1
0
1 Symbol Codeword
1
a2 0
a6 10
a1 110
a4 1110
a3 11110
a5 11111
 Huffman Coding Results

Avg. bit requirement/symbol:

H = -[(0.4)log2(0.4)+(0.3)log2(0.3)+(0.1)log2(0.1)
+(0.1)log2(0.1)+(0.06)log2(0.06)+(0.04)log2(0.04) C = 3/2.2 = 1.364
= 0.53+0.52+0.33+0.33+0.24+0.19 RD = (1- 1/1.364) = 0.267
= 2.14 bits/pixel Log2(x) = log10(x)x3.322
 Huffman Coding Example 2

Original Gray Level Probability Huffman code

(Natural Code)

g0: 002 0.2 010

g1: 012 0.3 00

g2: 102 0.1 011

g3: 112 0.4 1

➢ More frequently occurring fewer bits for the code

Huffman Coding Example – Entropy vs. Ave bpp

Example 2:
 Huffman Coding Example – Results

• In the example, we observe a 2.0 : 1.9

compression, which is about a 1.05
compression ratio, providing about 5%
redundant data compression.

• From the examples, we can see that the

Huffman code is highly dependent on the
histogram, so any preprocessing to
simplify the histogram will help improve
the compression ratio
 Huffman Coding

Properties :

❑ codes one symbol at a time

❑ Block code
❑ each symbol is mapped to a block of code
❑ Instantaneously decodable
❑ does not need to foresee the future codes while decoding
❑ Uniquely decodable
❑ any string of code symbol can be decoded in only one way
 Golomb Coding
The Golomb code of n with respect to m, denoted Gm(n),
is a combination of the unary code of quotient, floor[n/m]
and the binary representation of reminder (n mod m).

❑ Can only be used to represent nonnegative integers

inputs
❑ with exponentially decaying probability distributions
❑ can be optimally encoded
❑ using a family of codes.

❑ Computationally simpler than Huffman codes.

 Some Basic Compression Methods:
Golomb Coding
Given a nonnegative integer n and a positive integer divisor m  0,
Given
Given aa nonnegative
nonnegative integer
integer n
n and
and aa positive
positive integer
integer divisor
divisor m
m 
 0,
0,
Given a nonnegative
the Golomb code ofinteger
n withnrespect
and a positive integerGdivisor
to m, denoted ( n ), m  0,
constructed
the Golomb code of n with respect to m, denoted Gm ( n), constructed
the Golomb code of n with respect to m, denoted Gmm (n), constructed
theasGolomb
follows:code of n with respect to m, denoted Gm (n), constructed
as
as follows:
follows:
as follows:
Step 1. Form the unary code of quotient  n / m 
Step
Step 1.
1. Form
Form the the unary
unary code code of quotient  n
of quotient n // mm 
Step 1. unary
(The Form the codeunary code of
of integer q is definedas
quotient n /qm1s followed by a 0)
(The unary code of integer q is defined
(The unary code of integer q kis defined as q 1s followed as q 1s followed byby aa 0)
0)
Step2. Let k= log 2 m  , c = 2kk − m, r = n mod m, and compute truncated
(The unary code of integer q is defined as q 1s followed by a 0)
Step2.
Step2. Let
Let k=k= log
log 22 mm  ,, cc = 2 −
=k2 −mm,, rr =
=n n mod
mod m m,, and and compute
compute truncated
truncated
Step2. Let k= 
remainder r ' such
log m 

2 that , c = 2 − m , r = n mod m , and compute truncated
remainder r ' such
remainder r ' such that that
remainder r ' such rrthat
truncated to k -1 bits 0r c
r ' = r truncated to k -1 bits
truncated to k -1 bits 0
0 
 rr 
 cc
rr '' =
=rtruncated
r + c trunca to ted
k -1tobits rc
k bits 0otherwise
r ' =   rr + + cc trunca
truncated ted to
to kk bits
bits otherwise
otherwise

Step 3. Concatenate r + c trunca ted to k bits
the results of steps 1 and 2. otherwise
Step 3. Concatenate
Step 3. Concatenate the the results
results ofof steps
steps 11 and
and 2. 2.
Step 3. Concatenate the results of steps 1 and 2.
 Some Basic Compression Methods:
Golomb Coding

Step 1. Form the unary code of quotient  n / m  G4 (9) :

(The unary code of integer q is defined as
q 1s followed by a 0) 9 / 4  = 2,
the unary code is 110
Step2. Let k= log 2 m  , c = 2k − m, r = n mod m,
and compute truncated remainder r ' such that k= log 2 4  = 2, c = 22 − 4 = 0,
r = 9 mod 4 = 1.
r truncated to k -1 bits 0r c
r'= 
 r + c truncated to k bits otherwise r ' = 01

Step 3. Concatenate the results of steps 1 and 2. G4 (9) = 11001

 Some Basic Compression Methods:
Golomb Coding

Step 1. Form the unary code of quotient  n / m 

(The unary code of integer q is defined as
q 1s followed by a 0)

Step2. Let k= log 2 m  , c = 2k − m, r = n mod m,

and compute truncated remainder r ' such that G4 (7)?

r truncated to k -1 bits 0r c

r'= 
 r + c truncated to k bits otherwise

Step 3. Concatenate the results of steps 1 and 2.

 Golomb Coding

Fill the chart at home

n G1(n) G2(n) G4(n)

0
1
2
3
4
5
6
7 11111110 11101 1011
Loss Less Compression: Arithmetic Coding:

❑Variable length code

❑Error-free compression technique
❑Non block coding
– one to one correspondence between symbol and code
does not exist

– An entire sequence of source symbols (string of symbol)

is mapped to a single arithmetic number (code word)

– The code word itself defines an interval of real numbers

between 0 and 1.
❑ This coding can achieve theoretically higher compression
rates than Huffman codes
 Arithmetic Coding

• Maintains an interval between [0 1] based on the

probabilities of the symbol

A four symbol source

 Arithmetic Coding

❑Let we have a symbol sequence or message of a four

symbol source
a1a2 a3a3a4

❑ need to code the sequence with Arithmetic coding.

 Arithmetic Coding
At the start, interval [0,1) is
subdivided initially into four
regions based on the
probabilities
 Arithmetic Coding
a1a2 a3 a3 a4

Symbol a1, associated with

subinterval [0,0.2) is the first
message being coded and expanded
the full height in narrow range
 Arithmetic Coding
Symbol a1, associated with
subinterval [0,0.2) is the first
message being coded and expanded
the full height in narrow range

0.16

0.08

0.04
 Arithmetic Coding
Symbol a2, associated with
subinterval [0.04, 0.08) is expanded
the full height in narrow range again

0.16

0.08

0.04
 Arithmetic Coding
Symbol a2, associated with
subinterval [0.04, 0.08) is expanded
the full height in narrow range again

0.16 0.072

0.08 0.056

0.04 0.048
 Arithmetic Coding
Symbol a3, associated with
subinterval [0.056, 0.072) is
expanded the full height in narrow
range again

0.16 0.072

0.08 0.056

0.04 0.048
 Arithmetic Coding
Symbol a3, associated with
subinterval [0.056, 0.072) is
expanded the full height in narrow
range again

0.16 0.072 0.0688

0.08 0.056 0.0624

0.04 0.048 0.0592

 Arithmetic Coding
Now symbol a3, associated with
subinterval [0.064, 0.0688) is
expanded the full height in narrow
range again

0.16 0.072 0.0688

0.08 0.056 0.0624

0.04 0.048 0.0592

 Arithmetic Coding
Now symbol a4, associated with subinterval
[0.06752, 0.0688) is expanded the full height
in narrow final range. Any number within this
range can be use as message. Example: 0.068

0.16 0.072 0.0688

0.08 0.056 0.0624

0.04 0.048 0.0592

 Arithmetic Coding

The arithmetic code for the sequence

a1a2 a3a3a4
is 0.068

How to Decode it?

 Lempel-Ziv-Welch (LZW) Coding

◼ An error-free compression technique

◼ Removes spatial redundancy

◼ Assign fixed-length code words to variable length

sequences of source symbols

◼ It does not require any knowledge of probability of

occurrence

◼ LZW coding is used in the GIF, TIFF and PDF formats

 LZW Encoding – Images
❑ Images are scanned from left to right and from top to
bottom
❑ A codebook or dictionary containing the source
symbols to be coded is constructed on the fly
❑ For 8-bit monochrome images, first 256 words of the
dictionary are assigned to intensities 0,1,2,3,…,255.
 Lempel-Ziv-Welch (LZW) Coding
Coding Technique
❑ Generate a codebook/dictionary

– first 256 entries are assigned to gray levels

0,1,2,..,255.

– New gray level sequences not already in the

dictionary are assigned to a new entry.

– Example: for sequence “255-255” can be assigned to

entry# 256, the address following the locations
reserved for gray levels 0 to 255.
 LZW Encoding – Images

❑ The next time that two consecutive white pixels are

encountered, codeword 256 (the address of the location
containing 255-255) is used to represent them

❑ If 9-bit, 512-word dictionary is employed then two pixels

(16 bits) can be represented by 9 bits only
❑
 Lempel-Ziv-Welch (LZW) Coding

• Example
Consider the following 4 x 4 8 bit image Dictionary Location Entry

39 39 126 126 0 0
1 1
39 39 126 126 . .
255 255
39 39 126 126 256 -
39 39 126 126 511 -

Initial Dictionary
 Lempel-Ziv-Welch (LZW) Coding
39 39 126 126
39 39 126 126
39 39 126 126
39 39 126 126

Dictionary Entry
Location
0 0
1 1 Sequence= Concatenate(‘ ’, ‘39’) = 39
. .
255 255
256 -

511 -
 Lempel-Ziv-Welch (LZW) Coding
39 39 126 126
39 39 126 126
39 39 126 126
39 39 126 126

Dictionary Entry
Location
0 0
1 1 Sequence= Concatenate(‘ ’, ‘39’) = 39
. .
255 255
256 -

511 -
 Lempel-Ziv-Welch (LZW) Coding
39 39 126 126
39 39 126 126
39 39 126 126
39 39 126 126
Dictionary Entry
Location
0 0
1 1 Sequence= Concatenate(‘39’, ‘39’) = 39-39
. .
255 255
256 -

511 -
 Lempel-Ziv-Welch (LZW) Coding
39 39 126 126
39 39 126 126
39 39 126 126
39 39 126 126
Dictionary Entry
Location Address of ‘39’
0 0
1 1 Sequence= Concatenate(‘39’, ‘39’) = 39-39
. .
255 255
256 39-39

511 -
 Lempel-Ziv-Welch (LZW) Coding
39 39 126 126
39 39 126 126
39 39 126 126
39 39 126 126
Dictionary Entry
Location
0 0
1 1
. .
255 255
256 39-39

511 -
 Lempel-Ziv-Welch (LZW) Coding
39 39 126 126
39 39 126 126
39 39 126 126
39 39 126 126
Dictionary Entry
Location
0 0
1 1 Sequence= Concatenate(‘39’, ‘126’) = 39-126
. .
255 255
256 39-39

511 -
 Lempel-Ziv-Welch (LZW) Coding
39 39 126 126
39 39 126 126
39 39 126 126
39 39 126 126
Dictionary Entry
Location Address of ‘39’
0 0
1 1 Sequence= Concatenate(‘39’, ‘126’) = 39-126
. .
255 255
256 39-39

511 -
 Lempel-Ziv-Welch (LZW) Coding
39 39 126 126
39 39 126 126
39 39 126 126
39 39 126 126
Dictionary Entry
Location
0 0
1 1 Sequence= Concatenate(‘39’, ‘126’) = 39-126
. .
255 255
256 39-39

511 -
 Lempel-Ziv-Welch (LZW) Coding
39 39 126 126
39 39 126 126
39 39 126 126
39 39 126 126
Dictionary Entry
Location
0 0
1 1 Sequence= Concatenate(‘126’, ‘126’) = 126-126
. .
255 255
256 39-39

511 -
 Lempel-Ziv-Welch (LZW) Coding
39 39 126 126
39 39 126 126
39 39 126 126
39 39 126 126
Dictionary Entry
Location
0 0
1 1
. .
255 255
256 39-39
Sequence= Concatenate(‘126’, ‘39’) = 126-39
511 -
 Lempel-Ziv-Welch (LZW) Coding
39 39 126 126
39 39 126 126
39 39 126 126
39 39 126 126
Dictionary Entry
Location
0 0
1 1
. .
255 255
256 39-39
Sequence= Concatenate(‘39’, ‘39’) = 39-39
511 -
 Lempel-Ziv-Welch (LZW) Coding
39 39 126 126
39 39 126 126
39 39 126 126
39 39 126 126
Dictionary Entry
Location
0 0
1 1
. .
255 255
256 39-39
Sequence= Concatenate(‘39’, ‘39’) = 39-39
511 -
 Lempel-Ziv-Welch (LZW) Coding
39 39 126 126
39 39 126 126
39 39 126 126
39 39 126 126
Dictionary Entry
Location
0 0
1 1
. .
255 255
256 39-39
Sequence= Concatenate(‘39’, ‘39’) = 39-39
511 -
 Lempel-Ziv-Welch (LZW) Coding
39 39 126 126
39 39 126 126
39 39 126 126
39 39 126 126
Dictionary Entry
Location
0 0
1 1
. .
255 255
256 39-39
Sequence= Concatenate(’39-39’, ‘126’) = 39-39-126
511 -
 Lempel-Ziv-Welch (LZW) Coding
39 39 126 126
39 39 126 126
39 39 126 126
39 39 126 126
Dictionary Entry
Location
0 0
1 1
. .
255 255
256 39-39

511 -

Sequence= Concatenate(’126’, ‘126’) = 126-126

 Lempel-Ziv-Welch (LZW) Coding
39 39 126 126
39 39 126 126
39 39 126 126
39 39 126 126
Dictionary Entry
Location
0 0
1 1
. .
255 255
256 39-39

511 -

Sequence= Concatenate(’126-126’, ‘39’) = 126-126-39

 Lempel-Ziv-Welch (LZW) Coding
39 39 126 126
39 39 126 126
39 39 126 126
39 39 126 126
Dictionary Entry
Location
0 0
1 1
. .
255 255
256 39-39

511 -
 Lempel-Ziv-Welch (LZW) Coding
Encoding of

39 39 126 126
39 39 126 126
39 39 126 126
39 39 126 126
is
39 39 126 126
256 258 260 259 257 126

How to decode?
 LZW Decoding Dictionary Dictionary
Code word entry
0 0
Decoding of 1 1
| |
39 39 126 126 256 258 260 259 257 126 39 39
| |
is 126 126
39 | |
255 255
256 39-39
257 39-126
258 126-126
259 126-39
260 39-39-126
261 126-126-39
262 39-39-126-126
263 126-39-39
264 39-126-126
 LZW Decoding Dictionary Dictionary
Code word entry
0 0
Decoding of 1 1
| |
39 39 126 126 256 258 260 259 257 126 39 39
| |
is 126 126
39 39 | |
255 255
256 39-39
257 39-126
258 126-126
259 126-39
260 39-39-126
261 126-126-39
262 39-39-126-126
263 126-39-39
264 39-126-126
 LZW Decoding Dictionary Dictionary
Code word entry
0 0
Decoding of 1 1
| |
39 39 126 126 256 258 260 259 257 126 39 39
| |
is 126 126
39 39 126 | |
255 255
256 39-39
257 39-126
258 126-126
259 126-39
260 39-39-126
261 126-126-39
262 39-39-126-126
263 126-39-39
264 39-126-126
 LZW Decoding Dictionary Dictionary
Code word entry
0 0
Decoding of 1 1
| |
39 39 126 126 256 258 260 259 257 126 39 39
| |
is 126 126
39 39 126 126 | |
255 255
256 39-39
257 39-126
258 126-126
259 126-39
260 39-39-126
261 126-126-39
262 39-39-126-126
263 126-39-39
264 39-126-126
 LZW Decoding Dictionary Dictionary
Code word entry
0 0
Decoding of 1 1
| |
39 39 126 126 256 258 260 259 257 126 39 39
| |
is 126 126
39 39 126 126 | |
39 39 255 255
256 39-39
257 39-126
258 126-126
259 126-39
260 39-39-126
261 126-126-39
262 39-39-126-126
263 126-39-39
264 39-126-126
 LZW Decoding Dictionary Dictionary
Code word entry
0 0
Decoding of 1 1
| |
39 39 126 126 256 258 260 259 257 126 39 39
| |
is 126 126
39 39 126 126 | |
39 39 126 126 255 255
256 39-39
257 39-126
258 126-126
259 126-39
260 39-39-126
261 126-126-39
262 39-39-126-126
263 126-39-39
264 39-126-126
 LZW Decoding Dictionary Dictionary
Code word entry
0 0
Decoding of 1 1
| |
39 39 126 126 256 258 260 259 257 126 39 39
| |
is 126 126
39 39 126 126 | |
39 39 126 126 255 255
39 39 126 256 39-39
257 39-126
258 126-126
259 126-39
260 39-39-126
261 126-126-39
262 39-39-126-126
263 126-39-39
264 39-126-126
 LZW Decoding Dictionary Dictionary
Code word entry
0 0
1 1
Decoding of | |
39 39 126 126 256 258 260 259 257 126 39 39
| |
is 126 126
39 39 126 126 | |
39 39 126 126 255 255
256 39-39
39 39 126 126
257 39-126
39 258 126-126
259 126-39
260 39-39-126
261 126-126-39
262 39-39-126-126
263 126-39-39
264 39-126-126
 LZW Decoding Dictionary Dictionary
Code word entry
0 0
1 1
Decoding of | |
39 39 126 126 256 258 260 259 257 126 39 39
| |
is 126 126
39 39 126 126 | |
39 39 126 126 255 255
256 39-39
39 39 126 126
257 39-126
39 39 126 258 126-126
259 126-39
260 39-39-126
261 126-126-39
262 39-39-126-126
263 126-39-39
264 39-126-126
 LZW Decoding Dictionary Dictionary
Code word entry
0 0
1 1
Decoding of | |
39 39 126 126 256 258 260 259 257 126 39 39
| |
is 126 126
39 39 126 126 | |
39 39 126 126 255 255
256 39-39
39 39 126 126
257 39-126
39 39 126 126 258 126-126
259 126-39
260 39-39-126
261 126-126-39
262 39-39-126-126
263 126-39-39
264 39-126-126
 Compression Ratio using LZW Coding

❑ Suppose sequence 39-39-126 is replaced by 260

❑ In other words 3X8=24 bits are replaced by 9 bits

❑ Original image = 16X8 bits = 128 bits

❑ Compressed Image = 10X9 bits = 90 bits
❑ Compression Ratio = 128:90 = 1.42:1
 Recall
Run Length Encoding
• Image features
– All 256 gray levels are equally
probable → uniform histogram
(variable length coding can not
be applied)
– The gray levels of each line
are selected randomly so
pixels are independent of one
A computer generated
another in vertical direction
(synthetic) 8-bit image
– Pixels along each line are M = N = 256
identical, they are completely
dependent on one another in
horizontal direction

Spatial redundancy
110
 Recall
Run Length Encoding

• The spatial redundancy can be eliminated by using run-length pairs

(a mapping scheme)
• Run length pairs has two parts
– Start of new intensity
– Number of consecutive pixels having that intensity
• Example (consider the image shown in previous slide)
– Each 256 pixel line of the original image is replaced by a single
8-bit intensity value
– Length of consecutive pixels having the same intensity = 256

256 x 256 x 8
– Compression Ratio = = 128
[256 + 256] x 8
 Run Length Encoding (RLC)

1b 2w 1b 3w 4b 1w

• Uses run length pairs

– (0, 1), (1, 2), (0, 1), (1, 3), …

• Eliminates small spatial redundancies

• However, small runs results in expansion instead
of compression
113
 Run Length Encoding (RLC)

❑ Original image = 64X4 bits = 256 bits

❑ Compressed Image = 36X4 bits = 144 bits
❑ Compression Ratio = 64:36 = 16:9
 Run Length Encoding (RLE) in
Binary Image

❑ Run-length Encoding is also effective in case of binary images

❑can be represented as a sequence of runs only
– Example:

1b 2w 1b 3w 4b 1w

– Its representation: 121341

❑ Fix a way to determine the run values:
❑ Specify the value of first run
❑ Assume each row begins with a white run!!
 RLE – Binary Images
❑ Adjacent pixels in binary images are more likely to be
identical
❑ Scan an image row from left to right and code each contiguous
group (i.e. run) of 0s or 1s according to its length
❑ Establish a convention to determine the value of the run

❑ Common conventions are

❑ To specify the value of the first run of each row
❑ To assume that each row begins with a white run, whose run
length may in fact be zero
 RLE – Binary Images

❑ Example:

000001100000100000000101100000 (30 bits in a row)

❑ Specify the value of the first run of each row:

– 0 5 2 5 1 8 1 1 2 5: gray level of the first run is ‘0’ (black)

❑ Assume each row begins with a white run (bit ‘1’):

– 0 5 2 5 1 8 1 1 2 5: length of the first (white) run is 0
Use the convention that the first number corresponds to the
number of zeros (black) in a run
120
 Run Length Encoding in BMP

• Uses a combination of encoded and absolute mode

• Either mode can appear anywhere in the image
• Encoded mode:
– 2 byte RLC
– First byte: run length --Second byte: gray/color index
• Absolute mode:
– 2 byte RLC
– First byte: 0 --Second byte: as follows