基于人类视觉系统的扩散差错掩盖算法

Diffused Error Concealment Algorithm Based on Human Visual System

针对压缩视频传输过程中出现的信息丢失问题,提出了一种利用扩散方程来恢复丢失信息的差错掩盖算法.首先建立一种考虑时间与空间边界匹配的失真函数来估计丢失的运动矢量,对丢失信息进行初步恢复;然后利用人类视觉系统活动性掩盖效应特性,建立各向异性扩散方程,对丢失信息进行精细恢复;最后通过仿真实验对多个视频序列在离散丢失和连续丢失的情况下实现了差错掩盖.结果表明,文中提出的算法能有效去除重建图像的块效应,相对于传统算法,峰值信噪比平均增加3~5 dB,并且可获得更好的视觉效果.

In order to avoid the information loss during the compressed video transmission,an error concealment algorithm based on the diffusion equation is proposed to recover the lost information.In this algorithm,first,a distortion function considering the temporal and spatial boundary matching is presented to estimate the lost motion vectors and make an initial recovery of the lost information.Then,an anisotropy diffusion equation,which is constructed on the basis of activity-masking characteristics in human visual system,is used to make a fine recovery of the lost information.Finally,error concealments for different video sequences in discrete losing and consecutive losing cases are made by simulation.The results show that the proposed algorithm effectively reduces the blocking effect of the recovered images and achieves good visual effect,with a PSNR being averagely 3~5dB higher than that of the traditional algorithm.