块分类的双向估计加权边信息生成算法*

针对分布式视频编/解码中运动估计产生的预测重叠问题,提出了一种块分类的双向估计加权边信息生成算法。根据前后相邻关键帧对应块的相关性,将像素块分为保留块类和运动块类。对保留块类直接作保留处理;对运动块类中的像素块采用改进的双向运动补偿算法找到对应匹配块,并通过加权判决模块生成边信息。实验结果表明,在不增加编码端复杂度的情况下,算法有效提高了边信息质量。

Block-classified side information generation algorithm with bidirectional motion estimation and weight decision

In distributed video coding (DVC) system, the side information (SI) generation used motion estimation which brought an overlapped issue. This paper presented a block-classified side information generation algorithm which adopted an improved bidirectional motion estimation and weight decision. The pixel blocks were classified into two parts: the saving blocks and the moving blocks according to the correlation between the blocks in adjacent key frames. The saving blocks were retained directly. For the moving blocks, the matching blocks were found by an improved bidirectional motion estimation interpolation. Finally, generated the side information by using the weighted decision module. Simulation results show that the proposed algorithm can produce improvement in the quality of side information without increasing the complexity of the encoder.