Research On Wavelet-transform-domain Based Distributed Video Coding Scheme

Distributed Video Coding (DVC) is a new coding paradigm based on two major information theory results: Slepian-Wolf theory and Wyner-Ziv theory.Although the distributed video coding research dates back to the 1970's, the research work towards developing practical codec are more recent. In traditional video coding standards, the temporal correlation between adjacent frames is exploited through complex motion compensation/estimation work. Thus, because of high complexity encoder, traditional video coding standards can not provide perfect solutions in some new emerging applications which need low-complexity and low power consuming encoder. In order to solve this problem of complex encoder, the Distributed Video Coding is developed in these years. In this new solution, the temporal correlation is usually utilized at the decoder, so most of the computing complexity is shifted from encoder to decoder. Improved error resilience is another major functionality of this new video coding paradigm since the traditional encoder prediction loop and the associated error propagation do not exit anymore.In this thesis, the application background and theoretical foundation are presented firstly. Then the thesis gives out the fundamental implementation and state-of-art research results of DVC. Analysis about feasibility of wavelet transform and SPIHT algorithm using in DVC paradigm are given out later. Based on the above work, we propose a wavelet and SPIHT based distributed video coding codec -WDVC. In this codec, three types of correlation can be utilized in video coding: temporal, spatial and statistical correlations. In the proposed codec, the temporal correlation is explored at the decoder only through generating side information from two key frames to auxiliary decode Wyner-Ziv frame, then the encoder complexity is reduced significantly. The spatial correlation within the Wyner-Ziv frames is utilized by performing wavelet transform. As for the high-order statistical correlation, SPIHT algorithm is employed to explore this type of correlation. Worthy of mention is that the high-order statistical correlation is seldom considered in most of existing DVC systems, because of the difficulty of aligning the source and side information, we propose to use SPIHT in the proposed codec to explore the high-order statistical correlation among wavelet transform coefficients in order to further improve the coding efficiency of WDVC. Finally, the experimental performance and corresponding analysis are given out.