Research and developments of distributed video coding

Abstract

The recent developed Distributed Video Coding (DVC) is typically suitable for the applications such as wireless/wired video sensor network, mobile camera etc. where the traditional video coding standard is not feasible due to the constrained computation at the encoder. With DVC, the computational burden is moved from encoder to decoder. The compression efficiency is achieved via joint decoding at the decoder. The practical application of DVC is referred to Wyner-Ziv video coding (WZ) where the side information is available at the decoder to perform joint decoding. This join decoding inevitably causes a very complex decoder. In current WZ video coding issues, many of them emphasise how to improve the system coding performance but neglect the huge complexity caused at the decoder. The complexity of the decoder has direct influence to the system output. The beginning period of this research targets to optimise the decoder in pixel domain WZ video coding (PDWZ), while still achieves similar compression performance. More specifically, four issues are raised to optimise the input block size, the side information generation, the side information refinement process and the feedback channel respectively. The transform domain WZ video coding (TDWZ) has distinct superior performance to the normal PDWZ due to the exploitation in spatial direction during the encoding. However, since there is no motion estimation at the encoder in WZ video coding, the temporal correlation is not exploited at all at the encoder in all current WZ video coding issues. In the middle period of this research, the 3D DCT is adopted in the TDWZ to remove redundancy in both spatial and temporal direction thus to provide even higher coding performance. In the next step of this research, the performance of transform domain Distributed Multiview Video Coding (DMVC) is also investigated. Particularly, three types transform domain DMVC frameworks which are transform domain DMVC using TDWZ based 2D DCT, transform domain DMVC using TDWZ based on 3D DCT and transform domain residual DMVC using TDWZ based on 3D DCT are investigated respectively. One of the important applications of WZ coding principle is error-resilience. There have been several attempts to apply WZ error-resilient coding for current video coding standard e.g. H.264/AVC or MEPG 2. The final stage of this research is the design of WZ error-resilient scheme for wavelet based video codec. To balance the trade-off between error resilience ability and bandwidth consumption, the proposed scheme emphasises the protection of the Region of Interest (ROI) area. The efficiency of bandwidth utilisation is achieved by mutual efforts of WZ coding and sacrificing the quality of unimportant area. In summary, this research work contributed to achieves several advances in WZ video coding. First of all, it is targeting to build an efficient PDWZ with optimised decoder. Secondly, it aims to build an advanced TDWZ based on 3D DCT, which then is applied into multiview video coding to realise advanced transform domain DMVC. Finally, it aims to design an efficient error-resilient scheme for wavelet video codec, with which the trade-off between bandwidth consumption and error-resilience can be better balanced.