基于运动估计的整帧丢失错误掩盖算法

Error concealment algorithm for whole frame loss based on motion vector estimation

针对因信道错误引起的数据丢失而造成的整帧图像丢失的情况,提出了基于运动矢量估计的错误掩盖算法.假设图像局部区域的运动为线性运动,估计丢失块的运动矢量.利用该运动矢量指向的参考块的运动情况来验证估计运动矢量的可靠性.如果可靠,则使用该矢量重建丢失块;否则,采用边界匹配的方法重建丢失块.当参考帧中存在帧内宏块时,利用相邻宏块的运动信息估计其运动矢量.最后通过滤波来减轻重建数据块边界的块效应.实验表明,和已有算法比较,本文算法可以明显提高恢复图像的主观和客观质量.

An error concealment algorithm based on motion vector estimation was proposed for whole frame loss caused by a channel error. Lost motion vectors were estimated by the hypothesis of linear movement. Then the reliability of estimated motion vectors were evaluated by the motion vectors of reference blocks. Lost blocks related to reliable motion vectors were reconstructed. Other blocks were reconstructed by boundary matching method. The motion vectors were estimated by neighbor motion vectors when intra macroblocks existed in reference frames. A simple filter was employed to eliminate the block-edge effort across block boundaries. Experimental results show that the algorithm can achieve higher peak signal-tonoise ratio （PSNR） and better video quality compared with other existing algorithms.