一种高效低时延的多视点分布式视频编码

An Efficient Low-Delay Multi-View Distributed Video Coding

分布式视频编码是一种可以将编码端复杂度转移到解码端的视频编码方式.多视点分布式视频编码在视点内部大都采用运动补偿内插法来降低时间冗余,但是因为运动补偿内插法的解码顺序与视频顺序不一致,从而导致视频解码出现时延.因此,基于运动补偿外推法提出了一种高效低时延的多视点分布式视频编解码方法.此外,一般运动补偿外推法的重构质量要低于运动补偿内插法.在使用运动补偿外推法降低时间冗余的基础上,使用视差矢量内插法降低视点间冗余,经过融合算法提高重构质量.实验证明,所提出的这种高效低时延的多视点分布式视频编码,在重构质量相当的情况下,解码时间降低了0.3s.

Distributed video coding(DVC)is a video coding solution for shifting coding complexity from encoderto decoder. Multi-view distributed video coding(MDVC)mainly uses the motion compensation interpolation(MCI)to reduce the temporal redundancies inside the viewpoint. But the decoding order of MCI is inconsistent with the video sequence,which leads to decoding delay. Based on motion compensation extrapolation(MCE),an efficient low-delay MDVC was proposed. In addition,the general MCE reconstruction quality is lower than that of the MCI method. On the basis of the MCE,the disparity vector interpolation was employed to reduce the interview redundan-cies,while fusion algorithm was applied to improve the reconstruction quality. Results show that the decodingtime of the presented method is reduced by 0.3 s in the case of reconstruction quality of equal quality.