Avs Video Compression Motion Estimation Algorithm Research

AVS (Advanced Video Standard) is the video compress standard developed by our country. As our national intellectual property rights, these coding standards have higher coding efficiency and coding capability. Because temporal redundancy between the two frame in the video sequence, the video compression is possible by motion estimation. So as one part of AVS video compression, the research of motion estimation algorithm has important academic and realistic significance.By the further research of AVS motion estimation, the article has innovative research and design for the motion estimation algorithm of AVS video compression from three aspects. The task and innovation as fallows:1. Because the motion vector in video sequence has the character that it distributed more probably at horizontal and vertical direction than the other place with the same radius, on the basis of diamond search algorithm, the supposed algorithm in this article introduced a cross search cluster which including four pixels. The result show that new diamond search algorithm can overcome the shortage of traditional three step search which make the motion estimation trapped into local optimism, can reduce the power of residual matrix and improve the peak of ratio signal and noisy.2. According to the character of motion image in this article, the microblock in the smooth part and the microblock in the drastic part can be partitioned by C-mean. The mode of detached microblock in the drastic part can be decided by RDO (Rate Distortion Optimization). The mode decision of each microblock in the whole frame can be eliminated and the computational complexity can be decreased.3. Because of camera motion, video sequence has background motion. By global motion estimation, the parameters of camera can be estimated and be corded. After the parameters become code flow, they are transmitted onto channel. Then at the decoding end, they are decoded. The parameters of camera can be used to compute the motion vector of microblock in background by the mode of camera. The article choosed the accurate mode of camera and gain the six parameters by compute this mode. The pixels of foreground can be reduced by least square algorithm. At last, the accurate parameters can be gained.Summaries and prospects have also been put forward in the final of the artical.