EXIT-Based Side Information Refinement in Wyner-Ziv Video Coding

The accuracy of the side information (SI) is critical in the performance of distributed video coding algorithms. The side information is typically built at decoder based on the reconstructed data and on channel coding parity bits transmitted by the encoder. The optimal encoding rate is generally difficult to compute precisely due to the dynamics of video content with varying correlation. Effective methods for the refinement of imprecise side information are therefore important for improved decoding quality. In this paper, we propose to exploit the intrinsic property of channel coding algorithms in Wyner-Ziv video coding. The side information is refined via both the information-plane and the parity-plane bits, which rapidly increases the accuracy of refined SI. We use extrinsic information transfer (EXIT) chart analysis in order to estimate the variations of the mutual information in the iterative decoding. In particular, we characterize mutual information variations for punctured regular and irregular ratecompatible Low-Density Parity-check (LDPC) codes. Tracking the mutual information changes permits to decrease the coding rate of the information and parity bitstreams, while preserving the decoding quality. Simulations results confirm that our method improves on the decoding quality of recent distributed video coding algorithms, especially for high-motion sequences or at high coding rate regime.