End‐to‐end rate‐distortion optimisation of multiple description coding utilising multiple redundant pictures

Abstract

The rate-distortion optimisation of redundant picture-based multiple description coding is studied, and the minimisation of the end-to-end distortion is attacked by fully exploring the redundant allocation utilising multiple redundant pictures. The amount and relative encoding parameters of redundant pictures for each frame are adaptively decided for multiple redundant pictures, and this is different from the relevant existing description generation methods, which mainly focus on determining the amount of inserted redundancy with a fixed single redundant picture. The relationship of encoding parameters between each redundant picture of the same frame is derived according to the classification and formulation of relative distortion and rate. To accelerate the encoding process, the statistical properties between mismatch distortion, bit rate and encoding parameter are explored for the fast decision of redundant picture number. The experimental results show that the proposed method improves the overall end-to-end rate-distortion performance compared with the state-of-the-art multiple description coding method utilising redundant coding.