Optimization Of Video Encoding Based On MMX

At present, there are so many computing intensive programs in the software of multimedia and data communication.Although the amount of these programs are less than10%of the whole codes in the software system, it takes up90%of the calculating time during the program execution.Fortunately, these codes have common characteristics, such as much time occupying and many loops repeating.Code optimization is an important means to improving the performance of application program, SIMD, as a technology to improve the performance of application program dynamically, has been widely accepted by the industry community.In this paper it has introduced the principle of H.264encoding algorithm and the SIMD technology, elaborated the key techniques of H.264video encoding standard, and given deep analysis to the integral frame structure based on the source code, what is more, the functions that much too time-consuming in the H.264algorithm have been optimized by SIMD technology, thus to increase the efficiency of intra-coding.In view of the extensive application and strong practicability of video coding standard, along with the recent development of video compress coding and its characteristics, this paper mainly do some research on the real time application of H.264encoder based on X264, and its main work includes the following parts:Firstly, H.264video encoding standard as well as the main algorithm been used in it, and its key techniques has been researched by detail, and then the MMX and its feature have been summarized.Secondly, after the study of the X264source code, the integral frame structure of code stream of X264are proposed, and the main process in the structure has been discussed;and then further discussion has been given to the process which aims to optimize the more time-consuming modules in X264by use of MMX, which includes intra prediction, inter prediction, DCT transforming, filtering.Thirdly, in the comparison of the performance before and afteroptimization, not only at the angle of signal-to-noise, frame rate andcode rate, also from the aspect of the time-consuming, in short, the optimization results are checked out by multiple factors.After loading MMX, under the circumstance of signal-to-noise and code rate unchangeableframe rate has increased, and the time-consuming of analysis and encoding on macroblock has been decreased by a large margin, which achieves the optimization requirements.Finally, in the environment of VS2008, the codes are proved to debug through, the decoded code stream, the CIF video sequence, can transform real-time, which achieves the purpose of optimization.