Tutorial Sessions

Tutorial 01

Scalable Image and Video Coding - With and Without Wavelets
Speaker: Prof. Vikram Gadre, Dept. of Electrical Engineering, IIT Bombay


An important challenge in image and video coding has been to cater to a wide variety of encoder/ decoder capabilities. In more detail, the problem is as follows. The encoder often has a large amount of computational power and hardware/ software resources available to cater to a wide variety of requirements - some stringent, some less demanding. At the decoding end, there are many often many devices or terminals with widely varying capability and resource. One would like to design a source coding paradigm in which all these disparate requirements are simultaneously catered to, in the best possible manner. In doing so, each decoder must be able to do as well as possible, given its resources, and while catering to this diverse set of needs, the encoder must not need to recode the stream entirely for receivers with different capabilities. In other words, in going from one level of decoding capability to another, what must be provided is only the incremental information, so that graceful degradation is possible when one moves from a higher level of decoding capability to a lower one.

This problem has been addressed in the Scalable Video Coding standards which have only recently been formalized. The Scalable Video Coding (SVC) standards are an extension of the H.264/AVC standards. In the earlier standards that were known and accepted, scalability was present in greater or lesser measure. However, it was never fully exploited.

More or less in parallel to the development of different generations of video coding standards - MPEG-2, H.263, MPEG-4, and so on, wavelets and filter banks were studied by many researchers in signal and image processing. Wavelets are more than a set of functions - they represent the whole paradigm of multiresolution analysis and representation. In fact, wavelets have often been regarded as a natural tool for scalable representation, because of the inherent notion of incremental representation present at the helm of multiresolution analysis and filter bank theory. In fact, the JPEG 2000 standards use bi-orthogonal filter banks, which would lead to wavelet functions on iteration, for a possibly scalable representation of images. The literature on image and video coding is replete with several approaches to wavelet based scalable image and video coding. Though wavelets have provided scalability in the JPEG-2000 standards, they have not found their place of pride in the SVC standards.

This tutorial session is aimed at:

  • introducing the general notion of scalability in image and video representation, and introducing the types of scalability possible in general.

  • introducing the possibilities of scalable image and video coding based on wavelets and filter banks. The relevant fundamental concepts pertaining to wavelets and filter banks would be explained. A brief introduction to the JPEG-2000 standards would be provided.

  • introducing the Scalable Video Coding standards as a paradigm actually accepted by standards bodies. This would also mean discussing the types of scalability possible in video - temporal, spatial and quality.

    Tutorial Notes: (Acknowledgement)
    SVC - Part 1 (Download)
    SVC - Part 2 (Download)