Reducing correlation noise in Wyner-Ziv video coding

Abstract

Over the past few decades, Distributed Video Coding (DVC) has been considered as a compression paradigm suitable for applications which require simple encoding. Yet, the performance obtained with practical architectures is still far from the theoretical bound, mainly due to the inaccurate Side Information (SI) predicted at the decoder. The work presented in this paper tries to improve the correlation between the bit-planes of the SI and the corresponding bitplanes of the Wyner-Ziv (WZ) frame. The proposed algorithm uses the adjacent key frames to predict the quantization intervals for the SI and for the WZ frame. It then chooses whether the quantization module should use the floor or the ceiling operationsuch that theindices generated for the two frames differ by the smallest number of bits possible. Experimental results show that, for a given target quality, the proposed method can reduce the WZ bit-rate by up to 8.4% compared to the traditional coding schemes.