RD-OPT & Adaptive Huffman Coding

8/8/2019 RD-OPT & Adaptive Huffman Coding

http://slidepdf.com/reader/full/rd-opt-adaptive-huffman-coding 1/16

by

B.Nirmal



Introductiony In this project we proposed the efficient method of

image compression

y This technique is simple in implementation and

utilizes less memory

y An image of size 1024 x 1024 pixels, in which the

intensity of each pixel is an 8-bit quantity, requires 1

megabytes of storage space



y Digital storage for image processing applications falls in to

three principal categories

1) Short-term storage for use during processing

2)Online storage for relatively fast recall

3)Archival storage, characterized by Infrequent Access



JPEG Compressiony In lossy compression scheme,

The compression is achieved by dividing the image

in blocks of 8×8 pixels.

Applying a discrete cosine transform (DCT) on the

partitioned image.

Resulting coefficients are quantized less significant

coefficientsThe dc coefficient and ac coefficient values are

encoded separately



Fig 1.DCT based encoder/decoder processing steps



y Efficient quantization matrix has to be designed based

upon the rate and distortion specification

y In image compression the encoding scheme plays the

important role

y The number of bits that used to represent the coefficients

must be reduced.

y Adaptive Huffman coding is used to encode the

coefficients since it is single pass method



RD-OPT: Algorithm for Optimizing

DCT Quantization TablesIntroduction:

y A n Efficient algorithm for constructing DCT

quantization table with optimal rate -distortiontradeoffs for a given image

y It is used to compress the image at any desired

signal to noise ratio or compressed size



y other approach include

Psycho-visual model based quantization

Rate-distortion based quantization Stochastic Optimization quantization

y Our algorithm is based on the Rate Distortion

y The Key feature of this algorithm is optimize theQuantization table over wide range of rates anddistortions



Bit Rate for DCT Compressiony Rate is defined as the number of bits used to represent

per pixel

Rate= Size of compressed image

No of pixels in the image

y Increasing entries in the Q tends to decrease the rate

Rn(q)=(1/64)Entropy{(f[n]//q)},R(Q) R n(Q[n]).

n=0 to 63



RD-OPT Algorithmy

Input:An image I of W X H . Pixel range [0«..255]

yOutput:

R D-optimal DCT Quantization table Q

y Step 1: Gather DCT Statics for the image

y Step 2: use the statics to calculate R n(q) and Dn(q) for

each possible q

y Step 3: use dynamic programming to optimize R(Q) against

D(Q)



Adaptive Huffman coding

y Huffman coding requires K nowledge of probabilities

of source sequence . It is a two pass procedure

y In Adaptive Huffman coding no need of probabilities

of source symbol. So it is a single pass procedure

y It based on statics of the symbols already encountered.



Encoding Procedure



Decoding Procedure



Updating Procedure



Example



THANK YOU

RD-OPT & Adaptive Huffman Coding

Documents