Transform domain Wyner-Ziv video coding with refinement of noise residue and side information

Distributed Video Coding (DVC) is a video coding paradigm which mainly exploits the source statistics at the decoder based on the availability of side information at the decoder. This paper considers feedback channel based Transform Domain Wyner-Ziv (TDWZ) DVC. The coding efficiency of TDWZ video coding does not match that of conventional video coding yet, mainly due to the quality of side information and inaccurate noise estimation.

In this context, a novel TDWZ video decoder with noise residue refinement (NRR) and side information refinement (SIR) is proposed. The proposed refinement schemes are successively updating the estimated noise residue for noise modeling and side information frame quality during decoding. Experimental results show that the proposed decoder can improve the Rate- Distortion (RD) performance of a state-of-the-art Wyner Ziv video codec for the set of test sequences.