HEVC and 3D dual-tree discrete wavelet transform based multiple description video coding

Abstract

To improve the robustness of video transmission, a layered HEVC multiple description video coding (MDVC) based on 3D dual-tree discrete wavelet transform (3D DDWT) is proposed in this paper. The base layer of the coding structure consists of bitstreams coded by an HEVC with low bitrate, and then duplicated to each description. The error frames of the reconstructed base layer and the input video sequence are coded by a 3D discrete wavelet transform encoder to produce four coefficient trees with redundancy. After noise-shaping, these four trees are partitioned into two groups, forming the enhancement layer of diverse descriptions. As the directional selection property of 3D DDWT, the error matching problem can be solved without performing the motion estimation and motion compensation in each description, the complexity of the MDVC algorithm can also be reduced. Experimental results have shown that the proposed algorithm improves the error resiliency of HEVC while maintaining the coding efficiency.