Research And Implementation Of Motion Estimation Based On H.264

As an advanced video coding standard (Advanced Video Coding, AVC), H.264is declaredas the international standard in2003May, which is jointly developed by MPEG (MovingPicture Experts Group) and VCEG (Video Coding Experts Group). In addition to the sameadvantages with previous standards, it also has new technologies, which provides highercompression rate, better coding efficiency. In order to service HD applications preferably withH.264, improving the coding efficiency is one of the feasible measures. Based on the analysis ofH.264encoder, motion estimation and motion compensation technique can eliminate the videosequences of temporal and spatial redundancy to improve coding efficiency, it occupies mostoverhead of the entire coding process, So it is meaningful to do research on how to reduce themotion estimation time.This paper discusses the principle and new technologies of H.264coding standard; then itintroduces motion search algorithms used by H.264and compares their efficiency; finally, afterstudying the basic performance of UMHexagonS (Unsymmetrical-cross Multi-Hexagon-gridSearch) algorithm, it raises four improvements which are analysis of termination in advance,dynamic reference search window, non uniform search template and diamond search pattern, toachieve a new fast motion estimation algorithm based on UMHexagonS, it improves thesituation that it comes into a local minimum point status prematurely and intermittently,compared with original algorithm, meanwhile it saves the motion estimation time to ensure goodcoding efficiency, which makes H.264encoding applied in HD occasions better.Based on H.264standard test model JM10.2, it completes the implementation ofimprovements on VS2008coding environment. Through a variety of performance tests about6sequences with the different levels of activity, it finds that it reduces about18.32%of emotionestimation time on condition that it guarantees the same decoded video quality and the same bitrate basically. The new motion estimation algorithm raises the coding efficiency of the coder.