摘要
为了克服视频传输中因视频传输差错引起的视频质量下降 ,提出了一种视频差错掩盖的两步多权值边框匹配算法。该算法充分利用了受损宏块周围的宏块信息来估算受损宏块的运动向量 ,并首先通过预掩盖受损宏块 ,使视频图像质量初步改善 ;然后在此基础上利用两步多权值边框匹配算法评估受损宏块的候选运动向量 ;最后用产生最小边界匹配差值的候选运动向量的运动补偿块来替代受损宏块 ,以进一步改善图像质量。这样就基本上克服了边界匹配算法在物体边界对运动向量估算不准确以及大量连续 GOB或整帧丢失时 ,掩盖效果较差甚至无法掩盖等不足。通过基于 H.2 6 3+和 Internet的仿真实验表明 :该算法不仅能有效地抑制视频差错扩散 ,并能取得满意的差错掩盖效果。
Compressed video is vulnerable to transmission errors when transmitted over unreliable channels. In this paper, a two-step multi-weighted boundary-matching algorithm (TMBMA) for video error concealment is proposed to combat the transmission errors. With this algorithm, the motion vector (MV) of damaged block can be estimated by making full use of information of blocks around the damaged one. A pre-concealment is implemented to improve video frame quality to some degree and then the multi-weighted boundary-matching algorithm is used to evaluate every MV candidate of the damaged block. The MV candidate giving the minimum boundary-matching differences is selected to conceal the damaged block and video frame quality can be improved further. The disadvantages of side matching algorithm (SMA), such as incorrect MB displacement at object border and poor performance when one entire frame or plenty of consecutive GOBs are lost, are overcome with this algorithm. It is proved by H.263+ simulation results that satisfactory concealment performance can be achieved by effectively controlling the propagation of video errors.
出处
《中国图象图形学报（A辑）》
CSCD
北大核心
2004年第5期636-640,共5页
Journal of Image and Graphics
基金
铁道部科技研究开发项目 ( 2 0 0 3 X0 40 -A)
西南交通大学科学基金项目 ( 2 0 0 2 A0 4)
