Unit 5: Data Compression

Why compression?
 To reduce the volume of data to be transmitted (text, fax, images)
 To reduce the bandwidth required for transmission and to reduce storage
requirements (speech, audio, video)
Data compression implies sending or storing a smaller number of bits. Although many
methods are used for this purpose, in general these methods can be divided into two
broad categories: lossless and lossy methods.
 Storage Space and Coding Requirements
Storage Requirement
Uncompressed graphics, audio and video data require considerable storage capacity. For e.g. an
image represented in 640 by 480 resolutions with 8 bits per pixel require 300 kilobytes of
storage space. So, in order to reduce the storage requirement of the multimedia data it should be
compressed. The compression should be such that there should not be a loss of quality of the
data and hence compression should not reduce the information content of the data. For e.g. 90%
of the raw audio can be deleted without affecting the quality of the audio.
Bandwidth Requirement
Uncompressed data transfer requires greater bandwidth or data rate for its communication. If
the data is compressed there can be a considerable reduction in the bandwidth requirement for
the transmission of the data.
The following examples specify continuous media and derive the amount of storage required
for one second of playback:
An uncompressed audio signal of telephone quality sampled at 8 kHz and quantized with 8 bits
per sample requires 64 Kbits to store one second of playback and a bandwidth of 64 Kbits/sec.
An uncompressed stereo audio signal of CD quality is sampled at a rate of 44.1 kHz quantized
at 16 bits per sample require 705.6 * 103 bits to store one second of playback and the
bandwidth requirement is 705.6 * 103 bits/second.
 Storage Space and Coding Requirements
As mentioned above compression in multimedia systems is subject to certain constraints.
 The quality of the compressed data should be as good as possible.
 To make cost-effective implementation possible, the complexity of the technique used should be
minimal.
 The processing of the algorithm must not exceed certain time span.
In retrieval mode application, the following demands arise:
 Fast forward and backward data retrieval should be possible with simultaneous display.
 Random access to single images and audio frames of a data stream should be possible without
extending the access time more than 0.5 second.
 Decompression of data should be independent of other data units.
For both dialogue and retrieval mode, the following requirements apply
 The format should be independent of frame size and video frame rate.
 The format should support various data rate.
 There should be synchronization between the audio and video.
 Compression and decompression should not require additional hardware.
 The compression of data in one system of multimedia should ensure the decompression in the other
system.
STORAGE SPACE
Uncompressed graphics, audio and video data require considerable storage capacity
which in the case of uncompressed video is often not even feasible given today's
technology. Data transfer of uncompressed video data over digital networks requires
very high bandwidth to be provided for a single point to point communication. To
provide feasible and cost effective solutions, most multimedia systems handle
compressed digital video and audio data streams.

CODING REQUIREMENTS
Images have considerably higher storage requirements than text; audio and video have
even more demanding properties for data storage. Not only is a huge amount of
storage required, but the data rates for the communication of continuous media are
also significant.
 How is compression possible?
 Redundancy in digital audio, image, and video data.
 Properties of human perception
Redundancy:
Adjacent audio samples are similar (predictive encoding); samples corresponding to
silence (silence removal). In digital image, neighboring samples on a scanning line are
normally similar (spatial redundancy)
In digital video, in addition to spatial redundancy, neighboring images in a video
sequence may be similar (temporal redundancy) .

Human perception: Compressed version of digital audio, image, video need not
represent the original information exactly. Perception sensitivities are different for
different signal patterns. Human eye is less sensitive to the higher spatial frequency
components than the lower frequencies (transform coding).
DIFFERENCE BETWEEN LOSSLESS AND LOSSY DATA
COMPRESSION
Entropy, Source and Hybrid Coding
 Entropy Encoding
Entropy coding
An entropy coding is any lossless data compression method that attempts to approach
the lower bound declared by Shannon's source coding theorem, which states that any
lossless data compression method must have expected code length greater or equal to
the entropy of the source.
Run length encoding
This method replaces the consecutive occurrences of a given symbol with only one
copy of the symbol along with a count of how many times that symbol occurs. Hence
the names ‘run length’.
For example, the string AAABBCDDDD would be encoded as 3A2B1C4D.
A real life example where run length encoding is quite effective is the fax machine.
Most faxes are white sheets with the occasional black text. So, a run length encoding
scheme can take each line and transmit a code for while then the number of pixels,
then the code for black and the number of pixels and so on. This method of
compression must be used carefully. If there is not a lot of repetition in the data then it
is possible the run length encoding scheme would actually increase the size of a file.
 Entropy Encoding
Arithmetic coding
Arithmetic coding is a type of entropy encoding utilized in lossless data compression. Ordinarily,
a string of characters, for example, the words “hey” is represented for utilizing a fixed number of
bits per character.
Arithmetic coding yields a single code word for each encoded string of characters. The first step
is to divide the numeric range from 0 to 1 into a number of different characters present in the
message to be sent including the termination character and the size of each segment by the
probability of the related character.
Source coding
Source coding takes into account the semantics and the characteristics of the data. Thus, the
degree of compression that can be achieved depends on the data contents. Source coding is a
lossy coding process in which there is some loss of information content. For e.g. in case of
speech the speech is transformed from the time domain to frequency domain. In the
psychoacoustic the encoder analyzes the incoming audio signals to identify perceptually
important information by incorporating several psychoacoustic principles of the human ear. One
is the critical band spectral analysis, which accounts for the frequency regions than in lower ear’s
poorer discrimination in higher frequency regions. The encoder performs the psychoacoustic
analysis based on either a side chain FFT analysis or the output of the filter bank.
Example: pulse code modulation, delta modulation, fast Fourier transform.
 Hybrid Coding
This type of coding mechanism involves the combine use of both the source coding
and the entropy coding for enhancing the compression ratio still preserving the quality
of information content. The example of Hybrid Coding includes MPEG, JPEG,
H.261, DVI techniques.
 Major Steps of data compression

Preparation:
Preparation includes analog to digital conversion and generating an appropriate
digital representation of the information. An image is divided into blocks of 4*4 or
8*8 pixels and represented by a fixed number of bits per pixel.
 Major Steps of data compression
Processing: This involves the conversion of the information from the time domain to
the frequency domain by using DCT (discrete cosine transform). In the case of motion
video compression, inter frame coding uses a motion vector for each 8*8 block.

Quantization: It defines discrete level or values that the information is allowed to

take. This process involves the reduction of precision. The quantization process may
be uniform or it may be differential depending upon the characteristics of the picture.

Entropy Encoding: This is the lossless compression method where the semantics of
data is ignored but only its characteristics are considered. It may be run length coding
or entropy coding. After compression, the compressed video stream contains the
specification of the image starting point and an identification of the compression
technique may be the part of the data stream. The error correction code may also be
added to the stream. Decompression is the inverse process of compression.
 HUFFMAN ENCODING
Huffman coding is a lossless data compression algorithm. It is a lossless data compression
mechanism. It is also known as data compression encoding. It is widely used in image
(JPEG or JPG) compression. The idea is to assign variable-length codes to input characters,
lengths of the assigned codes are based on the frequencies of corresponding characters.
Huffman encoding idea are mention in following steps:
 It assigns a variable-length code to all the given characters.
 The code length of a character depends on how frequently it occurs in the given text or
string.
 A character gets the smallest code if it frequently occurs.
 A character gets the largest code if it least occurs.
Prefix rule: The variable-length codes assigned to input characters are Prefix Codes, means
the codes (bit sequences) are assigned in such a way that the code assigned to one character
is not the prefix of code assigned to any other character. The major steps involved in
Huffman coding:
 First, construct a Huffman tree from the given input string or characters or text.
 Assign, a Huffman code to each character by traversing over the tree.
 Huffman Tree
Step 1: For each character of the node, create a leaf node. The leaf node of a character
contains the frequency of that character.
Step 2: Set all the nodes in sorted order according to their frequency.
Step 3: There may exist a condition in which two nodes may have the same
frequency. In such a case, do the following:
1. Create a new internal node.
2. The frequency of the node will be the sum of the frequency of those two nodes that
have the same frequency.
3. Mark the first node as the left child and another node as the right child of the newly
created internal node.
Step 4: Repeat step 2 and 3 until all the node forms a single tree. Thus, we get a
Huffman tree
 Huffman Tree Example

Step 1: Make pairs of characters and their frequencies. i.e. (a,5), (b,9), (c,12), (d,13),
(e,16), (f,45)
Step 2: Sort pairs with respect to frequency, in this case given data are already sorted.
Step 3: Pick the first two characters and join them under a parent node. Add a new
internal node with frequency 5 + 9 = 14.
 Huffman Tree Example
Step 4:Repeat Steps 2 and 3 until, we get a single tree.
 Huffman Tree Example
Coding :Therefore, we get a single tree.Then,we will find the code for each character
with the help of the above tree. Assign a weight to each edge. Note that each left
edge weighted is 0 and the right edge weighted is 1.
 Huffman Tree Example

Here, no of character = 6
No of bits required to represent character = 2^3 i.e. code length will be 3.
Bits used after coding: 5*4 + 9*4 + 12*3 + 13*3 + 16*3 + 45*1 = 224
Therefore, (300-179)/300 = 40.3% of space is saved after Huffman coding.
 Huffman Tree Example
E.g. let the word to be encoded be AABABBEEEEFEFC
Length of string=14
No. of A =3
Probability of occurrence of A=3/14=0.214
No. of B=3
Probability of occurrence of B=3/14=0.214
No. of E=5
Probability of occurrence of E=5/14=0.357
No. of F=2
Probability of occurrence of F=2/14=0.142
No. of C=1
Probability of occurrence of C=1/14=0.01
 Huffman Tree Example

C= 111
F= 110
B= 10
A=01
E=00 Character Frequency Code Size
A 3 01 3*2=6
B 3 10 3*2=6
C 1 111 1*3=3
E 5 00 5*2=10
F 2 110 2*3=6
Total (5*8=40bits) 14bits 31 bits

Without encoding, the total size of the string was(14*8) 112 bits. After encoding the size
is reduced to 85bits(40+14+31)
 JPEG (Joint Photographic Experts Group)
The JPEG standard for compressing continuous –tone still pictures (e.g. photographs)
was developed by photographic experts working under the joint auspices of ITU, ISO and
IEC. JPEG is significant in compression because MPEG or the standard for motion
picture compression is just the JPEG encoding applied to each frame separately.
 There are some requirements of JPEG standard and they are:
 The JPEG implementation should be independent of image size.
 The JPEG implementation should be applicable to any image and pixel aspect ratio.
 Color representation itself should be independent of the special implementation.
 Image content may be of any complexity, with any statistical characteristics.
 The JPEG standard specification should be state of art(or near) regarding the
compression factor and achieved image quality.
 Processing complexity must permit a software solution to run on as many available
standard processors as possible. Additionally, the use of specialization hardware
should substantially enhance image quality.
 Why we need image compression?

Let us take an example, A typical digital image has 512x480 pixels. In 24 bit color (one
byte for each of the red, green and blue components), the image requires 737,280 bytes of
storage space. It would take about 1.5 minutes to transmit the uncompressed image over a
64 kb/second link. The JPEG algorithms offer compression rates of most images at ratios
of about 24:1. Effectively, every 24 bits of data is stuffed into 1 bit, giving a compressed
file size (for the above image dimensions) of 30,720 bytes, and a corresponding
transmission time of 3.8 seconds.
Steps in JPEG Compression
 Steps in JPEG Compression
Step 1: (Block/Image Preparation)
This step involves the block preparation. For e.g. Let us assume the input to be
640*480 RGB image with 24 bits/pixel. The luminance and chrominance component
of the image is calculated using the YIQ model for NTSC system.
Y=0.30R + 0.59G + 0.11B
I= 0.60R -0.28G -0.32B
Q= 0.21R-0.52G + 0.31B
Separate matrices are constructed for Y, I and Q each elements in the range of 0 and
255. The square blocks of four pixels are averaged in the I and Q matrices to reduce
them to 320*240. Thus the data is compressed by a factor of two. Now, 128 is
subtracted from each element of all three matrices to put 0 in the middle of the range.
Each image is divided up into 8*8 blocks. The Y matrix has 4800 blocks ,the other
two have 1200 blocks.
Steps in JPEG Compression
 Steps in JPEG Compression
Step 2: (Discrete Cosine Transformation)
Discrete Cosine Transformation is applied to each 7200 blocks separately. The output
of each DCT is an 8*8 matrix of DCT coefficients. DCT element (0,0)is the average
value of the block. The other element tells how much spectral power is present at each
spatial frequency.

Step 3: (Quantization)
In this step the less important DCT coefficients are wiped out. This transformation is
done by dividing each of the coefficients in the 8*8 DCT matrix by a weight taken
from a table. If all the weights are 1 the transformation does nothing however, if the
weights increase sharply from the origin, higher spatial frequencies are dropped
quickly.
 Steps in JPEG Compression
Step 4: (Differential Quantization)
This step reduces the (0,0) value of each block by replacing it with the amount it differs
from the corresponding element in the previous block. Since these elements are the
averages of their respective blocks, they should change slowly, so taking the differential
values should reduce most of them to small values. The (0,0) values are referred to as the
DC components; the other values are the AC components.
Step 5: (Run length Encoding)
This step linearizes the 64 elements and applies run length encoding to the list. In order to
concentrate zeros together, a zigzag scanning pattern is used. Finally run length coding is
used to compress the elements.
Step 6: (Statistical Encoding)
Huffman encodes the numbers for storage or transmission, assigning common numbers
shorter codes than uncommon ones. JPEG produces a 20:1 or even better compression
ratio. Decoding a JPEG image requires running the algorithm backward and thus it is
roughly symmetric: decoding takes as long as encoding.
Although JPEG is one algorithm, to satisfy the requirements of a broad range of still image
compression application. It has 4 modes of operation.
 Sequential DCT based
In this mode, 8x8 blocks of the image input are formatted for compression by
scanning the image left to right and top to bottom. A block consists of 64 samples of
one component that make up the image.
Each block of samples is transformed to a block of coefficients by the forward
discrete cosine transform (FDCT). The coefficients are then quantized and entropy
encoded.
 Sequential DCT based
Image preprocessing:
It basically involves the block preparation where the image samples are grouped into
8*8 pixels and passed to the encoder. Then Discrete Cosine Transformation is applied
to the blocks where the pixel values are shifted into the range [-128,127] with zero as
the center. Each of these values is then transformed using Forward DCT (FDCT).
DCT is similar to Discrete Fourier Transformation as it maps the values from the time
to the frequency domain.
Quantization
The JPEG application provides a table of 64 entries. Each entry will be used for the
quantization of one of the 64 DCT coefficients Each table entry is an 8 Each of the 64
coefficient can be adjusted separately. bit integer value (coefficient Qvu). The
quantization process become less accurate as the size of the table entries increase.
Quantization and DE quantization must use the same tables.
 Sequential DCT based
Entropy Encoding
During the initial step of entropy encoding, the quantized DC coefficients are treated separately
from the quantized AC-coefficient .
The DC-coefficient determines the basic color of the data units.
The DCT processing order of the AC coefficients involves the zigzag sequence to concentrate the
number of zeros.

AC coefficient are DCT coefficients for which the

frequency in one or both dimensions is non zero.
DC coefficient are DCT coefficients for which the
frequency in one or both dimensions is zero
 Sequential DCT based
JPEG specifies Huffman and arithmetic encoding as entropy encoding methods. However, as this
is lossy sequential DCT based mode, only Huffman encoding is allowed. In lossy sequential mode
the framework of the whole picture is not formed but parts of it are drawn i.e. sequentially done.
 Expanded/Progressive DCT based Model
This methods produce a quick low- resolution version of the image which is gradually
(progressively) refined to higher resolutions. This is particularly useful if the medium separating
the coder and decoder has a low bandwidth (e.g., a 14.4K modem connection to the Internet, in
turn providing a slow connection to a remote image database). The user can stop the download at
any time. This is similar to the sequential DCT multiple scans.
 Expanded/Progressive DCT based Model
For the expanded lossy DCT based model, JPEG specifies progressive encoding in addition to
sequential encoding. At first, a very rough representation of the image appears which is
progressively refined until the whole image is formed. This progressive coding is achieved by
layered coding. Progressiveness is achieved in two different ways:

 By using a spectral selection in the first run only, the quantized DCT coefficients of low
frequencies of each data unit are passed in the entropy encoding. In successive run the
coefficient of higher frequencies are processed

 Successive approximation transfers all of the quantized coefficients in each run, but single bits
are differentiated according to their significance. The most-significant bits are encoded first,
then the less-significant bits.
 LOSSLESS
The decoder renders an exact reproduction of the original digital image.
This mode is used when it is necessary to decode a compressed image identical to the
original. Compression ratios are typically only 2:1. Rather than grouping the pixels
into 8x8 blocks, data units are equivalent to single pixels. Image processing and
quantization use a predictive technique, rather than a transformation encoding one.
For a pixel X in the image, one of 8 possible predictors is selected (see table below).
The prediction selected will be the one which gives the best result from the a priori
known values of the pixel's neighbors, A, B, and C. The number of the predictor as
well as the difference of the prediction to the actual value is passed to the subsequent
entropy encoding.
 HIERARCHICAL
The input image is coded as a sequence of increasingly higher resolution frames. The
client application will stop decoding the image when the appropriate resolution image
has been reproduced. This mode uses either the lossy DCT-based algorithms or the
lossless compression technique. The main feature of this mode is the encoding of the
image at different resolutions. The prepared image is initially sampled at a lower
resolution (reduced by the factor 2n). Subsequently, the resolution is reduced by a
factor 2n-1 vertically and horizontally. This compressed image is then subtracted from
the previous result. The process is repeated until the full resolution of the image is
compressed.
Hierarchical encoding requires considerably more storage capacity, but the
compressed image is immediately available at the desired resolution. Therefore,
applications working at lower resolutions do not have to decode the whole image and
then subsequently reduce the resolution.
 MPEG COMPRESSION PROCESS
MPEG stands for Moving Picture Coding Exports Group. MPEG algorithms
compress data to form small bits that can be easily transmitted and then
decompressed. MPEG achieves its high compression rate by storing only the changes
from one frame to another, instead of each entire frame. The video information is then
encoded using a technique called Discrete Cosine Transform (DCT). MPEG uses a
type of lossy compression, since some data is removed. But the diminishment of data
is generally imperceptible to the human eye.
MPEG compression removes two types of redundancies:
Spatial redundancy:
 Pixel values are not independent, but are correlated with their neighbor both
within the same frame and across frames. So, to some extent, the value of a pixel is
predictable given the values of neighboring pixels.
 It is removed with the help of DCT compression.
 Temporal Redundancy
MPEG compression removes two types of redundancies:
Temporal redundancy:
 Pixels in two video frames that have the same values in the same location (some
objects repeated again and again in every frame).
 It is removed with the help of Motion compensation technique Macroblock
 It is the basic hierarchical component used achieving high level of compression.
 The key to achieving a high rate of compression is to remove much redundant
information as possible.
 Entropy encoding and Huffman coding are two schemes used for encoding video
information.
 MPEG takes the advantage of the fact that there exists a correlation between
successive frames of moving pictures.
 MPEG Compression
The MPEG compression algorithm encodes the data in 5 steps:
First a reduction of the resolution is done, which is followed by a motion
compensation in order to reduce temporal redundancy. The next steps are the Discrete
Cosine Transformation (DCT) and a quantization as it is used for the JPEG
compression; this reduces the spatial redundancy (referring to human visual
perception). The final step is an entropy coding using the Run Length
Encoding and the Huffman coding algorithm.
 Steps For standard MPEG Compression
Step 1: Reduction of the Resolution
The human eye has a lower sensibility to colour information than to dark-bright
contrasts. A conversion from RGB-colour-space into YUV colour components help to
use this effect for compression. The chrominance components U and V can be
reduced (subsampling) to half of the pixels in horizontal direction (4:2:2), or a half of
the pixels in both the horizontal and vertical (4:2:0).
Step 2: Motion Estimation
An MPEG video can be understood as a sequence of frames. Because two successive
frames of a video sequence often have small differences (except in scene changes),
the MPEG-standard offers a way of reducing this temporal redundancy. It uses three
types of frames:
I-frames (intra), P-frames (predicted) and B-frames (bidirectional)
 Steps For standard MPEG Compression
Step 2: Motion Estimation
The I-frames are “key-frames”, which have no reference to other frames and their
compression is not that high. The P-frames can be predicted from an earlier I-frame or
P-frame. P-frames cannot be reconstructed without their referencing frame, but they
need less space than the I-frames, because only the differences are stored. The B-
frames are a two directional version of the P-frame, referring to both directions (one
forward frame and one backward frame). B-frames cannot be referenced by other P-
or B-frames, because they are interpolated from forward and backward frames. P-
frames and B-frames are called inter coded frames, whereas I-frames are known as
intra coded frames.
The references between the different types of frames are realized by a process called
motion estimation or motion compensation. The correlation between two frames in
terms of motion is represented by a motion vector. The resulting frame correlation,
and therefore the pixel arithmetic difference, strongly depends on how good the
motion estimation algorithm is implemented. The steps involved in motion
estimation:
 Steps For standard MPEG Compression
Step 2: Motion Estimation
Steps involved in motion estimation are:
Frame Segmentation : The Actual frame is divided into no overlapping blocks (macro
blocks) usually 8x8 or 16x16 pixels. The smaller the block sizes are chosen, the more
vectors need to be calculated.
Search Threshold : In order to minimize the number of expensive motion estimation
calculations, they are only calculated if the difference between two blocks at the same
position is higher than a threshold, otherwise the whole block is transmitted.
Block Matching : In general block matching tries to “stitch together” an actual predicted
frame by using snippets (blocks) from previous frames.
Prediction Error Coding : Video motions are often more complex, and a simple
“shifting in 2D” is not a perfectly suitable description of the motion in the actual scene,
causing so called prediction errors.
Vector Coding : After determining the motion vectors and evaluating the correction,
these can be compressed. Large parts of MPEG videos consist of B- and P-frames as
seen before, and most of them have mainly stored motion vectors.
 Steps For standard MPEG Compression
Step 3: Discrete Cosine Transform (DCT)
DCT allows, similar to the Fast Fourier Transform (FFT), a representation of image
data in terms of frequency components. So the frame-blocks (8x8 or 16x16 pixels)
can be represented as frequency components. The transformation into the frequency
domain is described by the following formula:
 Steps For standard MPEG Compression
Step 4: Quantization
During quantization, which is the primary source of data loss, the DCT terms are
divided by a quantization matrix, which takes into account human visual perception.
The human eyes are more reactive to low frequencies than to high ones. Higher
frequencies end up with a zero entry after quantization and the domain was reduced
significantly.

Step 5: Entropy Coding

The entropy coding takes two steps: Run Length Encoding (RLE ) [2] and Huffman
coding [1].
These are well known lossless compression methods, which can compress data,
depending on its redundancy, by an additional factor of 3 to 4