The Research On The Intrafram Prediction Algorithm Technique Of H.264

H.264 also known as MPEG-4-Part 10, is a newly drawn up video frequency coding standard made unitedly by the video coding experts group (VCEG) of international telecommunication union (ITU-T) and moveing picture experts group (MPEG) of international standards organization (ISO/IEC) in 2002. H.26x has the features of better coding efficiency and network compaitibility, at the same time as inhering the advantages of previous. H.264 applied lots of new technology to improve the efficiency of video compression but in exchanges of the tremendous complexity of technology, this kind of coding is a limit of the applications of H.264 in the application fields. As such,the massive calculated quantities is the bottleneck for the extension of the application domain.In this thesis, a detailed analysis was made on the intra prediction coding of H.264, and a optimization of algorithm was executed for the avoidance of the huge calculated amounts generated in the process of intra prediction coding. Two parts of optimization strategy of intra prediction coding were created. In the first part, an improved three-step method was proposed in accordance with the 9 types of rediction algorithm of 4×4 luma block; and in the second part, three types of optimization algorithm based on different theories were proposed in terms of the improvements of the sentencing options between the modules of 4×4 luma block and 16×16 luma block.The improvements of the 9 prediction algorithm of 4×4 luma block were discussed firstly. The information of boundary blocks and DC prediction were added, then improvements were made in accordance with the theory that the 8 types of the mode excepting DC prediction have the directionality. There is an efficiency increasement after the improvements of algorithms.In the second part, research on the advance sentencing options of the two module of 4×4 luma block and 16×16 luma block was made. Three kinds of improved algorithm were proposed in this orientation, and every one has it's own advantage; The first one is a adaptive threshold algorithm based on the MAD value of each macroblock in every frame and the theory of mean value of those MAD value, the advantage is that the value will change with every frame in the vedio automatically. In the second kind, the theory of the combination of adjacent macroblock mode information and MAD value was employed, the advantage is the utilization of the dependencies of adjacent macroblocks, with resulted in a over smoothed transition between macroblocks of the compressed vedio. At last, the theory of the existed histogram translation method was utilized. An improvements on the threshold determination method were made, and the fomulas of double threshold and new threshold were proposed. The advantage is that it can convert the original mass calculated quantities to the simple and operation, and can change the former threshold determation method by setting up two thresholds on the altered fomula, therefore, the improved algorithm can adapted to the different vedio requirements under the different setup of the threshold.