The Research Of H.264 Encoder And The Realization On DSP

Nowadays the H.264 is advanced than any other former standards, at the same time TI company produces the high-performance general multimedia digital programmable processor TMS320DM642, with super computing power. It is a good solution to the higher algorithm complexity seriously influenced its application in real-time environment. It is to realize the H.264 encoding the ideal platform.The work of this paper is focused on the H.264 key algorithms and the DSP platform transplantation and optimize on the two aspects respectively.Firstly, the key technology used in the H.264 are analyzed, then for the intra-frame algorithm, the texture characteristic is adopted to judge the best intra-frame prediction mode before coding, which achieve the only purpose of a macro block mode during the coding process. Meanwhile, direction of intra prediction mode is adopted to achieve the purpose of a macro block mode. In addition, the study of the full search algorithm asymmetric cross-shaped multi-level hexagonal grid search algorithm (UMHexagonS),three-step search algorithm, the using of complex search patterns, the initial point of automatic forecasting and early stop strategy, a new fast motion estimation algorithm is proposed which to reduce the motion estimation time and improve the coding efficiency.Secondly, on the DSP platform, according to the structure of TMSDM642 and the H.264, this paper optimize on-chip memory, allocation of storage space to adjust the variable definitions and header file changes, Successful implementation of the DSP platform migration algorithm following Optimized using compiler options, data packaging technology, keyword, and other methods of the H.264 encoder is optimized.Through the optimization of the H.264 and the DSP, the encoder results in no major loss of coding circumstances, greatly increase encoding speed. Validated, the decoded video image, compared with the original image, the image quality is not significantly reduced.These method effectively reduce the running time of the encoder and provides a good reference for the future of software development based on this platform provides a good reference.