The Research Of Fast Motion Estimation Algorithm Based On H.264

The new generation video coding standard H.264uses some new technologies, such as the multi-mode motion estimation; intra prediction, multi-frame predictive, which greatly improves coding efficiency and network friendliness. The motion estimation is one of the key links as video encoding, motion estimation algorithm commonly uses multi-frame reference and tree structure motion compensation, and results in a sharp rise of the computational complexity, that severely restricted the speed of video encoding.In this paper, I study a variety of typical fast motion estimation algorithm, including the processes and key technologies of algorithms. Which focus on the analysis of the UMHS algorithms and the S-UMHS algorithm, the premise is to take full advantage of integer pixel motion estimation proportion of the various parts, optimize the redundant part of the S-UMHS algorithm search, and proposed improved S-UMHS algorithms. The algorithm applied early termination technology, optimization of the search process, the cross search and the hexagon search template to enhancing the overall performance of motion estimation, complete high-definition video surveillance coding. To further enhance the speed of motion estimation, the paper, using the GPU+CPU parallel structure and proposing parallel algorithms based on S-UMHS algorithm, designs parallel motion estimation part.The experiments show that, compared with the UMHS algorithms, the proposed improvement of the S-UMHS algorithms of search points reduced by80%, the motion estimation time is reduced by37-51%. And compared with the S-UMHS algorithms, the motion estimation time is reduced by6-16%. Improved S-UMHS algorithms maintain the rate-distortion performance, While maintaining a high compression efficiency, to meet the compression requirements of the large size image. The parallel algorithms based on S-UMHS algorithm greatly enhances the speed of motion estimation, and coding speed can reach74f/s, which is suitable for the real-time applications..In summary, the improved S-UMHS algorithm and parallel motion estimation algorithm has its own advantages. If we can combine the two advantages, which is good for the practical application.