Resaerch On Fast Motion Estimation Algorithm In H.264 Video Compression Coding

With the development of information technology, it has become indispensable part to get the video data in various stanard format through the network transmission in people’s daily life.The growing demands for the quality of video and the efficiency of video compression which has been pushing a new video coding standard to be generated and internation-alization.The H.264 standard proposed in 2003, inherited the previous codec structure of video coding standard, also introduced a number of advanced technologies, new technology makes the H.264 has a high compression ratio, but also brought lots of additional code load.Motion estimation is a key technology in the video compression encoding. It has a direct affect on the quality and efficiency of video coding. Motion estimation is also the most time-consuming part in the video compression encoding, according to statistics, the computational complexity of motion estimation about the whole coding process 80%. Therefore, research on effective motion estimation algorithm has a very important significances to improve the efficiency of video compression encoding, it is an efficient solution to improve video data compression efficiency. Firsty, this paper introduces the main frame, basic principle and key technology of H.264 Video Coding Standard, then further understanding the principles and strategies, and introduced several classic block matching motion estimation algorithm, in-depth study of the UMHexagon S algorithm, the weakness of the algorithm was optimized and improved.And on the basis of the algorithm design an adaptive hybird template search algorithm. UMHexagon S algorithm is the best performance in fast motion estimation algorithm of H.264 standard, but it still has many deficiencies. Firstly, the five kinds of starting search predict order of the algorithm is not optimal. Secondly, the 5×5 square search template of the algorithm dose not make full use of the vector distribution characteristics, lead to search points too much. Thirdly, it exists a lot of unnecessary search points in search for the macro block mode. The improved algorithm according to the initial prediction of MV become a possiblility size optimal point optimized forecast detection order of MV,it improves initial prediction accuracy; design an improved 5×5 full search template effectively reduces the search points; adds an early termination technique for sub macroblock, reduces a lot of unnecessary search. The simulation experimental results of JM platform show that the improved algorithm in the premise of effective ensure the quality of coding and efficiency, the motion estimation time an average reduction about 28%.Based on the combination with UMHexagon S algorithm and classical block matching motion estimation algorithm, an adaptive hybrid template motion estimation algorithm is proposed. Based on the space-time relativity of macro block, the algorithm achieved static block determination and ending the search static block directly; According to the type of movement, the algorithm adaptive selection of initial search point and search template, for a small motion block it will skip a big template coarse searching and into the small template fine searching directly. The experimental results show that the searching accuracy of the adaptive hybrid template motion estimation algorithm is close to UMHexagon S algorithm, but its searching speed is better than that of the existing fast motion estimation algorithm in H.264 standard.