Improving decoding speed for parallel distributed video coding architectures

Abstract

The Distributed Video Coding (DVC) paradigm is suitable for devices which have limited encoding capabilities. However, it is characterized by excessive decoding delays which compromise their application for time constrained services. This limitation can be mitigated by adopting parallel DVC architectures. Yet, the traditional Gray-code or binary-code representations have a non-uniform distribution of mismatch across bit-planes, resulting in uneven decoding times which hinder parallel decoding. This work proposes an alternative indexing scheme, where mismatch is distributed more uniformly amongst bit-planes and thus comparable decoding delays are expected, facilitating parallel implementations. This method reduces decoding time by up to 32% compared to architectures using simple parallel techniques, with a slight loss of 0.06dB in RD performance.