Chapter 8: Image Compression

Origin42

1.b. .i. Solution: 024

Given image matrix,

3 4 7

3 4 7

3 4 7

For 8-bit image, codebook will be 9-bit,

Dictionary
Entry
Location

0 0
1 1
… …
255 255

256 -
… -
511 -

Using LZW encoding process,

 Current Pixel Dictionary
Encoded Dictionary
Recognized Being Location
Output Entry
Sequence Processed (Codeword)

3 4 3 256 3-4

4 7 4 257 4-7

7 3 7 258 7-3

3 4 - - -

3-4 7 256 259 3-4-7

7 3 - - -

7-3 4 258 260 7-3-4

4 7 - - -

4-7 257

So, new codebook,

Dictionary Dictionary
Location Entry

0 0
1 1
… …
255 255

256 3-4

257 4-7

258 7-3

259 3-4-7

260 7-3-4

So, the output sequence is 3-4-7-256-258-257

1.b. .ii. Solution: 024
In my LZW coding process, it reduces spatial data redundancy. It assigns fixed-length code words
to variable length sequences of source symbols.

1.b. .iii. Solution: 024

Before compression,
Total matrix value 9
Each matrix value 8 bits
So, original image size = 9x8 =72 bits

After compression,
Total encoded output 6
Each encoded output 9 bits
So, compressed image size = 6x9 =54 bits

Compression ratio = 72/54 = 1.33 : 1

Yes, compression achieved.

5.a. . Solution: 024

(i)
Image compression is needed for:
● Data store at low storage
● Data transmission at minimum cost
(ii)
In image there are 3 types of data redundancies.
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.

(iii)
Entropy = 5.3 bits/pixel, means minimum 5.3 bits required to represent a pixel without losing any
data.
Compression ratio = (512x512x8) / (512x512x5.3) = 1.509 : 1
This is the maximum compression that can be expected.

5.b. Solution: 024

(i)
Given image matrix,

18 16 16

11 11 14

11 14 16

PDF,

Symbol Pixel Value Probability

a1 11 3/9 = 0.33

a2 14 2/9 = 0.22

a3 16 3/9 = 0.33

a4 18 1/9 = 0.11

Using Huffman coding process,

 a1 11
0
a3 0
1 1
6 0 1
1
a2
1
4

a4
1
8

Symbol Pixel Value Codeword

a1 11 0

a3 16 10

a2 14 110

a4 18 111

(ii)
Original image entropy, H = -[(0.33)x log2(0.33) + (0.22)x log2(0.22) + (0.33)x log2(0.33) + (0.11)x
log2(0.11)] = 1.7578 bits/pixel

(iii)
Huffman coding is a Coding Redundancy reduce technique.

Average bit requirement per symbol, Lavg = 0.33x1 + 0.22x3 + 0.33x2 + 0.11x3 = 1.98 bits/pixel

As, H < Lavg. So, this compression is lossless.

(iv)
Compression ratio = 8/1.98 = 4.04 : 1
So, compression was achieved.
 Enigma41

2.a. Solution:

2.b. Solution: 024

(i)
Given, image pixels,

f(x, y) f’(x, y)

1 4 3 3 6 5
=>
2 3 4 4 5 6

2 4 5 4 6 7

Total 9 pixel values. Each pixel value is increased by 2.

Row=M=3, Column=N=3

Erms = sqrt((1/9) x (2^2) x 9) = 2

(ii)

Gm(n) = G4(9)
m =4
n =9

Step 1:
floor(n/m) = floor(9/4) = 2
So, unary value followed by 0 = 110

Step 2:
k = ceiling(log2(m)) =ceiling(log2(4)) = 2
c =2^k - m = 2^2 - 4 = 0
r = n mod m = 9 mod 4 = 1
r’ = 01 ; [r+c truncated to k bits as 0<=r<c is not accomplished]

Step 3:
Golomb code for G4(9) = “ unary value followed by 0 ” + “ r’ ” = 11001
 Recursive40

5.a. Solution: Rabab 039

(i) Image compression is needed for:
● Low data storage
● Data transmission at lower cost

(ii) There are 3 types of data redundancies: Coding, spatial, irrelevant information

(iii) Arithmetic Coding is a lossless compression technique that encodes data by creating a code string
which represents a fractional value on the number line between 0 and 1

(iv)
(v)
5.b. Solution:
 Prototype39
1.

Solution:

Solution:

Return38
5.

Solution:
Solution: Rabab 039
(i) Total pixels in image = 4*8 = 32

Symbol Pixel value Probability

a2 95 4/32 = 0.125

a3 169 4/32 = 0.125

a1 21 12/32 = 0.375

a4 243 12/32 = 0.375

Entropy = - [0.125log(0.125) + 0.125log(0.125) + 0.375log(0.375) + 0.375log(0.375) = 0.796 (Amar

1.81127 ashe ~ ID 077 log2 hobe)

(ii) Using Huffman Coding:

Symbol Pixel value Probability Huffman codeword

a2 95 4/32 = 0.125 0
 a3 169 4/32 = 0.125 10

a1 21 12/32 = 0.375 110

a4 243 12/32 = 0.375 111

(iii) Lavg = (0.125x1) + (0.125x2) + (0.375x3) + (0.375x3) = 2.625

Compression ratio, C = (4x8x8) / (4x8x2.625) = 8 : 2.625 = 3.0476 : 1

(iv) RD = 1 - 1/C = 1 - 1/3.0476 = 0.671

(v)

Solution:
Given in Origin42, 1(b)

Malware37
1.
Solution:
A grayscale image with most of the pixel values being the same is a spatial redundancy. This can be
overcome by many compression techniques involving spatial redundancy, most notably LZW coding.

The image is as follows:

39 40 60 80 100

39 40 60 80 100

39 40 60 80 100

39 40 60 80 100

39 40 60 80 100

Generating a 9-bit codebook:

Location Entry

0 0
… …
255 255

256 -
… …
511 -

Using LZW:

Current Dictionary
Pixel being Encoded Dictionary
Recognized Location
processed output entry
sequence (codeword)

39

39 40 39 256 39-40

40 60 40 257 40-60

60 80 60 258 60-80

80 100 80 259 80-100

100 39 100 260 100-39

39 40 - - -

39-40 60 256 261 39-40-60

 60 80 - - -

60-80 100 258 262 60-80-100

100 39 - - -

100-39 40 260 263 100-39-40

40 60 - - -

40-60 80 257 264 40-60-80

80 100 - - -

80-100 39 259 265 80-100-39

39 40 - - -

39-40 60 - - -

39-40-60 80 261 266 39-40-60-80

80 100 - - -

80-100 39 - - -

80-100-39 40 265 267 80-100-39-40

40 60 - - -

40-60 80 - - -

40-60-80 100 264 268 40-60-80-100

Location Entry

0 0
… …
255 255

256 39-40

257 40-60

258 60-80

259 80-100

260 100-39
 261 39-40-60

262 60-80-100

263 100-39-40

264 40-60-80

265 80-100-30

266 39-40-60-80

267 80-100-39-40

268 40-60-80-100

269 -
… …
511 -

Before compression,
Total outputs = 25
Bit size = 8
So, image size = 25*8 = 200

After compression,
Total outputs = 13
Bit size = 9
So, image size = 13*9 = 117

Compression ratio = 200 : 117 = 1.709 : 1

So, compression is achieved.

COREi36
3.

Solution:
Solution:

Solution: