H.264视频码流时域误码掩盖算法的改进

Improvements of Temporal Error Concealment Schemes for H.264 Video Streams

新一代视频压缩标准H.264/AVC中推荐的解码时域掩盖算法是基于宏块的,预测矢量集仅包括了当前帧中的相邻矢量.为此,提出了基于分块的、利用加权边界匹配准则的精细运动补偿时域掩盖算法.该算法改进了匹配准则、扩展了预测矢量集、提高了运动矢量恢复的准确性,并采用一种自适应时/空掩盖算法,将时域/空域误码掩盖有效地结合起来.算法的改进均基于解码端,不改变H.264原有的码流结构,具有广泛的应用性.3GPP/3GPP2移动IP信道传输的仿真实验证明,该改进算法具有比H.264推荐的相应算法更强的错误鲁棒性.

Temporal error concealment schemes suggested in the newly composed video coding standard H. 264 are based on macroblocks, and only include spatial neighboring motion vectors in the candidate MV set. A method of subblock-based refined motion compensated concealment with weighting boundary match criteria was proposed. The candidate MV set is adaptively expanded, which improves the accuracy for motion vector recovery. An adaptive spatial/temporal estimation method with low complexity is used to effectively combine temporal and spatial concealment. The promotions are focused on decoder side, without any structure changes on H. 264 video encoder and transport technology. They can be widely employed in multiple video devices. The simulation of transmission over 3GPP/3GPP2 mobile IP channels shows that the modified concealment schemes provide improvements in error resilience over those suggested in H. 264.