Click to edit Master subtitle style

Data Compression

Chapter 1 Introduction

Anokhi Shah
4/15/12

1 1 Anokhi

Data Compression

Data compression is art or science of representing information in a compact form form is created by identifying and using structure that exist in the data.
Structure could be any Pattern, Redundancy

Compact

or Model.
Example

Data AAABBBCCC character is repeated 3 times)

(Every

Fewer bits information

required

to

represent
Anokhi

22

4/15/12

Examples
ZIP

File (Image) (Sound) type English

JPEG MP3 SMS

Information : Info About : Abt Are: r You : u

Code
4/15/12

33

Anokhi Shah

Anokhi

Advantages & Disadvantages

the consumption of expensive resources such as Hard disk, Bandwidth transformation of Data compressed data can only be understood if the decoding method is known by the receiver data must be decompressed to be viewed or heard.
44

Compressed

Anokhi Shah

4/15/12

Anokhi

1.1 Compression Techniques

is a compression algorithm that takes an input X a representation generate the reconstruction Y. Xc to

generates By

Compression algorithm, we mean Compression as well as Reconstruction algorithm. Xc

Anokhi Shah

Reconstruc
4/15/12

Anokhi

1.1.1 Lossless Compression

is identical to X loss of information

No

Lossless

compression is reversible so that the original data can be reconstructed exactly used for applications that cant tolerate any difference between the original and reconstructed data compression algorithms usually exploit statistical redundancy
4/15/12 66

Generally

Lossless
Anokhi Shah

Anokhi

Lossless Compression Examples

Compression

It is very important that reconstruction is

identical to original

Very For

small difference can result statements with different meanings

in

a Bank application consider the sentences Do not send money and Do now send money

Radiological

/ Satellite image
77

If data of any kind are to be processed or

enhanced later to yield more information,

Anokhi Shah

Anokhi

1.1.2 Lossy Compression

is different from X loss of information

Some

Compressed Obtains

data cannot be recovered or reconstructed exactly much higher compression ratio than is possible with lossless compression used for applications that can afford to have difference between the original and reconstructed data
4/15/12 88

Generally

Anokhi Shah

Anokhi

Lossy Compression Examples

/ Image

a person viewing a picture or television

video scene might not notice if some of its finest details are removed or not represented perfectly

Sound

/ Speech

Both parties of the telephone conversation

can ignore certain loss of information.

99 Anokhi Shah 4/15/12

Anokhi

1.1.3 Measure of Performance

compression algorithm can be evaluated in a number of different ways

algorithm algorithm

relative required

complexity to

of

the the

implement

How fast the algorithm performs on a

single machine
Amount of compression How closely the reconstruction resembles

the original

10 10

Anokhi Shah

4/15/12

Anokhi

Compression Ratio
Number

of bits required to represent the data before compression to the number of bits required to represent the data after compression an image requires 65536 bytes for storage, After compression it requires 16384 bytes ratio is 4:1

Suppose

Compression Ratio

can be expressed as the reduction in the amount of data required as a 11 percentage of the size of the original 11 data. Anokhi 4/15/12 Anokhi Shah

Compression Rate
The If

average number of bits required to represent a single sample we assume 8 bits per byte (pixel), the average number of bits per pixel in the compressed representation is 2 we would say that rate is 2 bits per pixel.

Thus

12 12 Anokhi Shah 4/15/12

Anokhi

Distortion / Fidelity / Quality

difference between the original and the reconstruction is called the distortion when you talk about the difference between reconstruction and original its called fidelity/ quality Y-X = =Distortion Fidelity or Quality, X-Y

But

i.e

When

fidelity or quality of a reconstruction is high, the difference 13 between the reconstruction and original 13 is small Anokhi 4/15/12 Anokhi Shah

1.3 Modeling and coding

scheme depends on various factors like characteristic of the data. compression technique that works on text would not work on image because both have different kind of data compression development is in 2 phases
Modeling Coding
14 14 4/15/12

Data

Algorithm

Anokhi Shah

Anokhi

1.3 Modeling and coding

modeling phase we try to extract information about any redundancy that exist in the data and describe the redundancy in the form of a model coding phase, a description of the model and a description of the data differs from a model are encoded generally using a binary alphabet difference between data and model is referred as residual
15 15 4/15/12

In

The

Anokhi Shah

Anokhi

Example 1.2.1
On

page no 7 We are required to store the binary representation of the numbers. Max no is 21, which will require 5 bits because binary of 21 is 10101 So rate is 5 bits per sample By plotting the data as seen in figure 1.2 (Pg no 7) ,the data falls on a straight line with an equation Xn= n + 8 Xn is the model and En is the residual
16 16 Anokhi Shah 4/15/12

Anokhi

Residual sequence consist of 3 numbers which can be represented in binary using 2 bits
n 1 2 3 4 5 6 Xn Xn = n+8 9 9 11 10 11 11 11 12 14 13 13 14 En=XnXn 0 1 0 -1 1 -1
4/15/12

-1 00 0 01

We need to make use 1 10 of structure that exist in data to predict the value of each element in the sequence and then encode the residual.
17 17

Anokhi Shah

Anokhi