Based On Tms320dm642 H.264 Encoding Optimization

From video phone, meeting phone,DVD, flow media to digital TV broadcasting, digital video has been used more and more widely, and in which setting standard is playing an important impelling role. Mepg-X set by ISO/IEC and H.26X set by ITU-T are two main standard systems in the current field of video compression. The former has been established standard in the circle of Radio & Tv Broadcast, and the latter has been widely used in the field of video phone and meeting TV.How to develop new standard by using the newest technology to meet the demand of new services is becoming urgent. As a result, the two standard organizations ISO/IEC and ITU-T cooperated again to set new video codec standard H .264/MPEG-4 Part 10. The standard defines so many new techniques such as multi frame motion compensation, 1/4-sample motion compensation, integer DCT transform and so on. The bit rate can be reduced more greatly than any other video compression standard. And it is possible to transmit the more and the higher quality video in fixed bandwidth.We realized the H.264 coder on EVM based on DSP TMS320DM642 made by TI recently. Firstly in the paper, we detail H.264 standard after getting the knowledge about video compression. My classmate joined in the design and improvement of coder flow chart based on software JM73, migrated and optimized the JM73 software to the TMS320DM642 EVM. TMS320DM642 is based on the kernel of TMS3200 C64**. The hardware resourses provide nessesary qualifications for high quality. How to use the hardware resourses and development tools adequately to make the code reach the expected function is the most important task in this paper. Depending on the constructure of TMS320DM642 and unique characteristic of H.264 application program, we optimized whole structure and DCT/IDCT, quazitation and vlc which cost most of coding time.After the relative optimization of program structure and modules, we found that the efficiency of DCT/IDCT, quazitation and vlc is improved greatly than before by testing the time, that is to say, the efficiency of entire coder is improved finally. At the same time, we get a lot of real data for future optimization.