Transform_Coding - DATA COMPRESSION / Transform Coding 3....

Info iconThis preview shows pages 1–2. Sign up to view the full content.

View Full Document Right Arrow Icon
DATA COMPRESSION / Transform Coding ESKISEHIR OSMANGAZI UNIVERSITY, DEPT. OF ELECTRICAL & ELECTRONICS ENG. 1 3. Transform Coding Entropy coding techniques described in previous sections work by removing redundancy from the input data. The definition of the redundancy inherently calls for the definition of the messages or symbols in the data. Whatever the implied message set is, entropy coding techniques preserve all the bits and pieces of the input data. This is an obvious necessity for compression of the most data types. This is sometimes equivalent to carrying a bucketful of water and not wasting a single drop of it. For some data type we can do with a bottle of it and it is not worth to save/carry every drop of it. We would have some allowance to spill some around while working with such data type. Computer images and digital voice are among such types. Some loss in the image data can be tolerable considering the anticipated gains in terms of the compression ratios. The key phrase in lossy compression techniques is the balance between the fidelity of the output data and the compression ratio. Some applications impose an upper limit on the amount of -compressed- data and therefore the compression ratio is determined accordingly. The quality of the output is a secondary issue and accepted as it is. Real time video compression-transmission under the pressure of bandwidth requirements of the channel is an example to such applications. Transmission of each and every bit of information embedded in such images is too expensive or just not feasible. Entropy coding compression techniques alone would not be much help since the average best entropy compression ratio on image data is proven to be about 1:3. The worst thing is that we just can not stop or slow down the incoming -supposedly- real time data. However we might be able to satisfy low bit rate requirement at the cost of some loss in the quality of the output video. For most of the time, what to remove from the input data is not obvious. The most naive action is to reduce the bits per image elements, which actually helps a lot for some image types. It would be luxurious to carry 32-bit color images for architectural schematics where only a couple of bits could actually do. However, when we talk about removing redundancy from images, it is obvious that we mean much further than eliminating those extra bits. The objective here would be to represent the image or image sets using minimum number of bits on the average and introducing minimum amount of distortion to images while doing that. In order to achieve this, one has three redundancies to play with in the images; Coding redundancy Psycho-visual redundancy Interpixel redundancy Eliminating the extra bits is in the first category. Previous section dealt with such coding techniques, so they are not going to be repeated here. Psycho-visual redundancy is related to how human eye and brain collaborate to make us fill in the blanks at the points where they believe there should not be.
Background image of page 1

Info iconThis preview has intentionally blurred sections. Sign up to view the full version.

View Full DocumentRight Arrow Icon
Image of page 2
This is the end of the preview. Sign up to access the rest of the document.

This note was uploaded on 12/12/2009 for the course BTU image codi taught by Professor Osmannurimuhat during the Spring '09 term at Kadir Has Üniversitesi.

Page1 / 9

Transform_Coding - DATA COMPRESSION / Transform Coding 3....

This preview shows document pages 1 - 2. Sign up to view the full document.

View Full Document Right Arrow Icon
Ask a homework question - tutors are online