Rate controlled redundancy-adaptive multiple description video coding
Image Communication
| Hi-index | 0.00 |
To overcome drawbacks of the low complexity multiple description (MD) video coding method which we presented in a previous paper, we propose an improved MD video coding method in this paper. The previous method consists of three integral components, namely GOB alternation, motion information duplication, and low quality macroblock update, in which the first two components establish a basic MD coding framework, and the last one makes redundancy in a controllable way. The method has very good performance and a relatively lower complexity. However, the method can not reach as large a redundancy region as some other MD video coding methods. One cause is that duplicating motion in both the descriptions introduces much redundancy for video sequences with slower motion. Another cause is that with limited quantization parameter range specified by the video coding standards, the lowest redundancy region which low quality macroblock update can reach is limited too. This paper addresses the problems by introducing two additional coding modes for a low quality inter macroblock, both modes take into account strategies a decoder will take for error concealment. The two modes take only 2 bits each to indicate in the bit stream, so redundancy introduced by motion duplication and low quality macroblock update is greatly reduced. A rate distortion criterion is used for choosing which mode to use for a particular low quality inter macroblock. Simulation results show, with some increment in computational complexity, the new method achieves even better performance than the previous approach both in terms of redundancy rate distortion, and in packet loss scenario. The new method can also reach a much lower redundancy region.