【摘要】 为进一步提高分布式视频编码的WZ帧重构质量,在分析现有算法基础上,提出一种WZ帧重构算法。从实际应用角度出发,对Laplacian和Cauchy分布相关噪声模型进行分析,指出其各自特点,在此基础上将二者混合使用,通过设定阈值对残差系数进行分类,分别对应生成边信息质量高和边信息质量低的情况,生成高质量边信息采用Laplacian分布进行建模,反之采用Cauchy分布表示相关噪声,分别推导WZ的条件期望值作为WZ重构值。实验中选用参数0.5,其结果表明,该算法可以显著提高WZ帧质量和系统率失真性能,适用于运动剧烈视频序列和大GOP情况。更多还原

【Abstract】 To further improve the quality of WZ frame reconstruction,a WZ frame reconstruction algorithm was proposed after the existing algorithm was analyzed.From the perspective of practical application,Laplacian and Cauchy distribution correlation noise model were analyzed,and their respective characteristics were pointed out.Two correlation noise model were mixed to use,and the residual coefficient was classified by setting threshold which respectively corresponded to the quality of generated side information.Laplacian distribution was used as correlation noise model for high quality side information,Cauchy distribution was used as correlation noise model for low quality side information,and the condition expectation of WZ was computed as the reconstructed WZ value.Experiment parameter was fixed to 0.5.The results show that the proposed algorithm can significantly improve the quality of WZ frame quality and system rate distortion(RD)performance,especially it is suitable for dramatic motion video sequences and large GOP sizes.更多还原