Layered Tree-based Multiple Description Coding

In this paper, a layered tree-based multiple description coding framework, called the layered spanning multiple description coding (LSMDC), is proposed to exploit the advantages from both multiple description coding and layered coding. With a tree-structured codec system, the wavelet coefficients are encoded to form one base layer and multiple enhancement layers followed by the use of an X-tree codec individually. Each layer consists of the codewords from one or more coding passes. To generate multiple descriptions, codewords are partitioned into several sub-layers according to the importance of related coefficients. The data structure of the proposed LSMDC is organized in a layered spanning tree graphical expression, and each description is formed by combining codes along a path from the root of the tree to one of its leaves. Simulation results have shown that the performance of the proposed LSMDC codec outperforms existing state-of-the-art multiple description codecs in terms of less center and side distortions