Motion Estimation Algorithm Analysis And Optimization In H.264/AVC Standard

International video compression standard H.264 is very popular in video compression community for its simple, estimated effective, and coding efficiency. H.264 video coding standard can save as much as about half of the bit stream than past standards under the premise of guaranteeing the same image quality. However, the introduction of several new coding techniques makes the computational complexity of the encoding and decoding increase dramatically. Of those which based on block matching motion estimation account for the entire video encoding operations as much as 80%. This high volume of complex operations has become the bottleneck of video compression coding for real-time application, Therefore, motion estimation algorithm of low complexity and high accuracy has become a research hotspot in video compression technology in recent years.This article firstly introduces the work of Block-Matching Motion Estimation in detail, and describes several classic fast algorithms for block-matching characteristics as well as their advantages and disadvantages. Then, by analyzing the motion characteristics of video sequences, we propose a big-small cross parallel search and multi-pattern self-adaptive search algorithm for motion estimation. Because this algorithm makes full use of characteristics that static and quasi-stationary block are in the majority in video sequences, plus the use of multi-end strategy, so the proposed algorithm greatly simplifies the motion estimation search process and improves H.264 coding speed. At the same time, self-adaptive multi-pattern search optimize motion intensity partition method, and ensure the accuracy of motion estimation and coding efficiency.In order to verify the validity of this algorithm, we applied this algorithm in H.264 reference model for system simulation experiment, and compare with two classical algorithm (FS algorithm and UMHexgonS algorithm). By analysis of experimental data, experimental results show that, in the premise of no significant change in image quality and algorithm stability, the algorithm save motion estimation time 15.88% on average than UMHexgonS. with the trade-off of the motion estimation time, PSNR values, and bit-rate, the proposed self-adaptive pattern-based double cross-searching algorithm perform superior to UMHexgonS generally with more robust, more conducive to real-time applications.