一种有效降低误差扩散的数据分割方法

A Data Partitioning Method for Reducing Error Propagation

提出一种有效降低误差扩散的数据分割方法．该方法改进了H．264划分子流的做法，采用当前编码宏块被错误解码时对下一编码帧产生的误差扩散程度来判别该宏块的优先级，以此划分不同的传输子流．在给定网络带宽和丢包率的情况下，采用率失真优化的码率分配算法对不同传输子流分别施以不同的信道编码保护力度，以保证在相同传输条件下，优先级高的宏块被正确接收的概率最大化，使重建视频流的质量达到最优．该方法实现简单，对宏块的分类更加合理，有效降低了由于包丢失引起的误差扩散效应，改善了重建视频的质量．实验结果表明在相同的带宽和丢包率条件下，本文方法比H．264方法提高视频接收质量大约0．3～0．6dB．

In this paper,a novel data partitioning method is presented, aiming at reducing error propagation for video streaming. The proposed method modifies the approach defined in H. 264, differentiating the importance level of every encoded macroblock by using the degree of the error propagation caused by the loss of the macroblock and then forming separate sub-bitstreams. Given the network bandwidth and packet loss rate, a bit rate allocation algorithm based on rate-distortion optimization is presented to protect individual sub-bitstreams at different levels, which enhances the possibility of correctly receiving the macroblocks of high level and improves the reconstructed quality of video streaming. The proposed method is easy to implement. Most important, the method effectively reduces the error propagation as packet loss happens. The simulations show that the proposed method gains 0.3-0.6dB than H. 264 under the same channel conditions.