Research And Parallel Implementation Of Motion Estimation Algorithm In H.264 Video Encoding

H.264/AVC video encoding standard is the new generation video compression standard, which was developed by VCEG(Video Coding Experts Group) and MPEG(Moving Picture Expert Group) together. This encoding standard has many improvements over the previous standards and the compression ratio has reached 1.5 to 2 times than the previous. High compression efficiency, low latency and high quality network transmission is a bright spot for H.264/AVC. Multi-reference frames, multi-block modes and other advanced video encoding technology of H.264/AVC make the H.264/AVC encoding standard performance much better than the previous encoding standards. However, because of its high encoding efficiency, the complexity of encoding standard is higher than the previous standards. It becomes a hot spot and focus of video encoding to study the various modules of encoding standard. The study shows that nearly 50%of video encoding time is spent in motion estimation, so the research of motion estimation algorithms optimization has become a top priority,this paper introduces H.264/AVC video encoding standard and the basic principles related to the standard. The paper makes a detailed description of the Enhanced Predictive Zone Search(EPZS) and discusses some shortcomings of the algorithm, for example, predictive subset is not accurate and the search template is too complicated. This article makes three improvements in the EPZS algorithm about SI subset, S4 subset and search templates and then conducts the experiments using video sequences which the intensity is different. The results show that the speed of encoding raises 8.1% while the quality of video sequence is only negligible decreased.After improving the algorithm of motion estimation, this article suggested a procedural framework of parallel execution for different mode selection and search templates of EPZS, and then the paper do experiments with some video sequences which intensity is different. The results show that the speed of encoding has improved 37% while the video quality decreases a little compared with the regular encoding.