Study On H.264 Fast Motion Estimation Algorithm

H.264 is the ITU-T's Video Coding Experts Group and ISO/IEC for Moving Picture Coding Experts Group was set up Joint Video Team development of new generation video coding standard. The use of a number of new technologies, such as multi-mode prediction, multiple reference frame motion estimation, integer transform, CABAC, loop deblocking filter, the result has high coding efficiency of H.264.In addition to the high coding efficiency, H.264 standard, the introduction of the network adaptation layer encapsulation of the code stream, packaging after the stream of network adaptability. In addition, for the transmission of the packet loss and bit error phenomenon, H.264 offers more and better anti-error strategy. But the H.264 coding efficiency, is based on the cost of high computational complexity in exchange, and its encoding complexity beyond the H.263/MPEG-4 more than tenfold. Handling capacity of the existing platform, simply can not do real-time H.264 encoding. Therefore, the encoder must be optimized, reducing the computational complexity.This paper mainly studies the fast motion estimation algorithm for H.264. First of all, reserch the fast inter prediction algorithm, respectively, to optimize the two directions from the following: in the choice interframe macroblock partition mode, the first part of the model calculations according to judge the candidates in advance mode group, and then from the candidate model to select the best mode of calculation of the group; in split block motion estimation, the first detection of local typical position of the blocks based on test results to predict the movement area, and then there is basis for the use of the corresponding search templates to search.Secondly, also the energy-based texture analysis of fast mode decision algorithm is studied. The algorithm block in advance the level of detail and texture direction analysis, according to mode selection block level of detail to narrow the scope of models based on texture direction of further screening, thereby greatly increasing the speed pattern search, will control the performance loss can be ignored range.Finally, the verification algorithm to optimize analyzed and concluded.