Вы находитесь на странице: 1из 16


-- Muktinath Vishwakarma 7th Semester (CSE) RGCER, NAGPUR

Thanks to:
Prof. H.R. Turkar Sir.

Data compression is the art of reducing the number of bits needed to store or transmit data.

Compression reduces the size of a file: To save space when storing it. To save time when transmitting it. Most files have lots of redundancy.

1: Lossless Compression
-- Shannon-Fano Algorithm -- Huffman Coding -- LZW Compression

2: Lossy Compression
-- Transform Coding.

1: DCT (Discrete Cosine Transform)

2: KCT (Karhunen-Laeve Transform)

-- Distortion Measures -- The Rate Distortion theory -- Quantization 1: Uniform Scalar Quantization 2: Non Uniform Scalar Quantization -- Transform Coding

1: DCT
2: KCT

Mathematical Quantity Specify how close an approximation is to its original,
using some distortion criteria.

Where, Row is Mean square error (MSE), Xn is input SNR (O/MSE), PSNR(Peak/MSE)

data sequence, Yn is reconstructed data sequence, N is length of the data sequence.

Always involves a tradeoff between rate and distortion. Rate is the average number of bits required to represent
each source symbol.

The tradeoff between rate and distortion is represented

in the form of rate distortion function R(D).

Heart of Any Lossy scheme Aim to reduce number of distinct
values to a much smaller set. 1: Uniform Scalar Quantization 2: Non Uniform Scalar Quantization

Partition the domain of inputs values into equal
spaced intervals, except possibly at the two outer intervals. quantizer's decision boundaries.

The endpoint of partition intervals are called the Output/value corresponding to each interval is taken
to be the mid point of the intervals.

length of each interval step size (delta triangle) It is of two types.

1: Midtread ( 0, Odd no of o/p level )
2: Midrise( (0), Even no of o/p level )

If input source is not uniformly distributed. It may be inefficient. Increasing the number of decision levels within the
regions where the source is densely distributed can effectively lower granular distortion. number of decisions levels, we can enlarge the region in which the source is sparsely distributed.

In addition, Without having to increase the total

Such Non Uniform quantizers thus have non
uniformly defined decision boundaries.

Coding Vectors is more efficient than coding scalar We need to group block of consecutive samples from
source input into vectors.

The rationale behind transform coding:

If Y is the result of a linear transform T of the input vector

X in such a way that the components of Y are much less correlated, then Y can be coded more efficiently than X.

If most information is accurately described by the first few

components of a transformed vector, then the remaining components can be coarsely quantized, or even set to zero, with little signal Distortion.


A widely used transform coding technique, is able to perform decorrellation of the input signal in a data-independent manner. Because of this it has gain tremendous popularity. Definition of DCT:

Fundamental of Multimedia, Ze-Nian Li, Mark S.

http://mattmahoney.net/dc/dce.html#Section_6 http://en.wikipedia.org/wiki/Data_compression http://www.ics.uci.edu/~dan/pubs/DataCompression.h



http://www.data-compression.com/index.shtml http://www.cs.princeton.edu/~rs/AlgsDS07/20Compres