The Mpeg2 To Avs And H.264 To Avs Transcoding Study

During the recent decades, the international Video Coding Standard developed the utility of digital video technology like Digital TV, Network Multimedia Stream, Network Video Meeting etc. The main video coding standards that are widely used nowadays include H.26x series, MPEGx series, and AVS. Among all these standards, MPEG2 is the most widely used one in Software, Hardware Design and Multimedia Data Storage domain. But in some bandwidth limited application areas like Wireless Transmission, Network Video Meeting, MPEG2 is not the best choice. Compared to MPEG2, AVS has higher coding efficiency and owns the key video coding patent. In conclusion, the transcoding from MPEG2 to AVS is one viable and effective resolution in video technology application area.In this dissertation, the video transcoding technology background and actuality are first discussed. Then we introduce the key coding technology of MPEG2 and AVS to put forward the pixel cascade transcoding framework and a fast transform domain transcoding framework from MPEG2 to AVS. Meanwhile a fast I frame transform domain transcoding algorithm is put forward in this paper. The transcoding framework is implemented on PC palteform using MPEG2 TM source code and AVS rm52f source code. According to analyzing the time cost of all function modules, we found that the inter frames' MV re-estimation module and the half/quarter pixel interpolation module in I/P frames cost the most time in the transcoding process. A MV re-using algorithm between inter frames and SSE code are adopted to optimize the project. Meanwhile I frmaes' prediction modes in AVS encoder are decided by DCT coefficients from MPEG2 decoder. The test result of three resolution stream verifies that the algorithm put forward in this paper accelerate the transcoding process with little picture distortion and transcode CIF .m2v stream to .avs stream in real time. The "H.264 to AVS transcoding project" is discussed in the fifth chapter. The prediction modes mapping algorithm and S transformation are put forward in this project.