分布式多视点视频编码容错传输研究

在分布式多视点视频编码(Distributed Multi-view Video Coding, DMVC)数据传输过程中,编码方式不同导致K帧与WZ帧受信道误码影响也不相同,因此本文提出了一种DMVC整体容错传输框架,针对K帧及WZ帧特性设计了不同的容错保护传输方案并进行有效融合。针对于K帧的容错传输问题,首先根据左、右相邻视点的对应K帧,利用DIBR算法产生的空间边信息对丢失块进行初始修复；然后根据K帧同一视点内的相邻已解码帧,生成它的时间参考帧,对K帧的丢失块进行重修复。针对WZ帧的容错传输问题,本文提出了基于不等错误保护(Unequal Error Protection, UEP)的编码算法,根据不同频带的各个比特面的重要性不同,对低频带、高比特面进行更加合理的码率分配,在不增加编码端复杂度的前提下提高了WZ帧的误码容错性能。实验结果表明：在K帧和WZ帧均出现丢包的情况下(丢包率为5%~15%),相比K帧采用传统的帧内错误隐藏加WZ帧采用参考文献码率的算法,本文方案对视频序列重建图像的BD-PSNR平均提升了2.39~4.68 dB,且随着丢包率的增加,提升效果更加显著。

Research on Error Resilient Transmission for DMVC

In the typical DMVC system, the different coding scheme results in different error influence to transmission error in wireless channel for K frames and WZ frames. In this paper, a new error resilient DMVC coding scheme is proposed for better error resilience, in which different coding methods are designed specifically for K frame and WZ frame. For K frames, firstly, the losing blocks led by transmission error are initially repaired with the spatial side information of adjacent viewpoints applying DIBR algorithm. Then, the spatial side information for decoded K frame is generated based on the adjacent decoded WZ frames in the same view. For WZ frames on the other hand, a new coding scheme based on UEP is designed, a new rate allocation algorithm regarding the different level of significance for different frequency bands, bitplanes is proposed. As a result the robustness of WZ frames is improved without increasing the complexity of the encoder. Experimental results show that the error resilience performance have been improved significantly for both the K frames and WZ frames. More specifically, compared with conventional error concealment algorithms for K frames and rate allocation algorithms for WZ frame, the BD-PSNR for different video sequences can be improved from 2.39 dB~4.68 dB, at the packet loss ratio of transmission from 5%~15%. In the meantime, the higher of the channel error, the more improvement can be achieved.