Implementation And Optimization Of H.264 Encoder Based On DSP

Transmitting the video of high-quality through the narrow bandwidth and adapting to various networks are always pursuits of people, so a new international video coding standard was released on Mar 2003, and was jointly being made by ITU-T Video Coding Experts Group(VCEG) and ISO/IEC Moving Picture Experts Group(MPEG),named as joint video team(JVT). This standard was known as ITU-T Recommendation H.264 and as ISO/IEC 14496(MPEG-4) Part 10. Compared with other prior video coding standards, such as MPEG-2, MPEG-4 ASP and H.263, about 50% bits can be saved and more adaptive to various networks with design of 2-layers. H.264 has been gotten much attention from research area and industry all through, but the high-complexity of H.264 has always been a factor of restriction of application mostly, so to study the implement and optimization of H.264 based on DSP is significant.In this paper, based on the research of standard and reading of magazines, the key technologies of H.264 are introduced first, and the technologies to enhance the efficiency of encoding are emphasized on; then the fast multi-mode decision algorithm especially fast multi-mode motion estimation (FMME) algorithm is provided, and some experiments are present. Based on these, the baseline encoder on DM642 is implemented. From the point of view of programmer, the methods of optimization on DSP are analyzed in detail, these parts, which based on research of experiments; at last, the performance of H.264 network camera is given, and from the results of experiment, this encoder, which can realize real-time coding has a good value in practice. Therefore what has been studied should be a promotion to make H.264 work in practice and in wide use.