Research On Motion Estimation Algorithm Based On H.264/AVC Encoder

H.264/AVC(Advanced Video Coding) is an advanced video encoding standard proposed together by ISO/IEC MEPG(Moving Picture Experts Group) and ITU-T VCEG(Video Coding Experts Group) in2003. Compared with previous widely used technical standard, the new generation standard has advantages on high compression ratio、low requirement for transmission bandwidth and high quality of image, which is available for many kinds of application scenario. With the network Infrastructure rapidly developed in recent years, real time high-definition video (HD video) application explode, so H.264/AVC can be used in this field. After the carefully analysis of the standard encoding algorithm, we can find that H.264/AVC applies many new technologies such as multiple reference frames selection, variable block-size modes, fraction-pixel motion estimation, rate control strategy and so on, which is able to improve the coding efficiency. But on the other hand, all of these new methods significantly increase the complexity of algorithm, therefore we must optimize the encoder in order to realize the application of H.264/AVC in the real time video system.Research shows that motion estimation takes more than70%H.264/AVC encoding time in all of the algorithm module, thus the optimization of the motion estimation can determine the performance of the encoder because it is the core technology. The efficiency of encoder gains significant improvement by motion compensation technique and motion estimation, which play an important role in eliminating the time redundancy in video sequences. So how to design effective fast motion estimation algorithm becomes a hot area in the video encoder field.This paper began with an introduction of the basic concept and core technologies in the H.264/AVC encoding standard, meanwhile lists all former standards and compares their performance, then analyzes the principle of the motion estimation and discusses several common motion estimation algorithm. This paper mainly pays attention to one of the most important fast motion estimation algorithm UMHexagonS(Unsymmetrical-cross Multi-Hexagon-grid Search),and carries out a detailed study of the algorithm, then shows its advantages and disadvantages. Finally, some improvements are introduced in the UMHexagonS to decrease the encoding time. They are the partition of the motion estimation search area, optimization of coarse search template and hexagon template and the introduction of new threshold value.The experiments result show that the new improved UMHexagonS takes less encoding time but maintain the same bit rate and PSNR compared with the FS(Full Search) and UMHexagonS. The new algorithm is able to save about20.33%Motion Estimation time and the biggest change of PSNR is only0.03dB. The bit rate increases only0.51%, so the new algorithm effectively lowers complexity and increases the real time performance of the encoder.