Wyner-Ziv Successive Refinement of Video and Rate Distortion Analysis

In Wyner-Ziv video coding system, motion estimation efficiency is much lower than that in conventional video coding system because current frame is not available when doing motion estimation. In this paper, we propose a successive resolution refinement algorithm to improve motion estimation efficiency. Based on our rate distortion analysis, we derive optimal down-sample ratio for two-stage successive resolution refinement system. We also analyze the performance of multistage case, and find that it approaches the performance of ideal motion compensated Wyner-Ziv video coding system, with at most 2.17 dB loss in PSNR. Experimental results demonstrate the correctness of the analysis and show that the proposed method out-performs original bit-plane refinement scheme up to 2.5 dB, with much lower complexity.