The Implementation And Optimization Of AVS Video Decoder On DM642

AVS is the second national audio video coding standard of China with independent intellectual property rights. As the primary part to compete with other international standards, AVS plays a key role in obtainning the independence in digital audio video industrial. It came into effect in 1^st March, 2006 as national standard and became one of the international IPTV standards in 2007.Comparing to other international coding standards such as MPEG-2 and H.264/AVC, AVS has many advantages such as trivial patent cost, lower complexity and well-defined applications. With the government's strong support of high-tech industries, AVS industry chain becomes mature and has a good prospective future.It is difficult to implement the real-time video coding on hardware because of the complexity of digital video coding and its high needs about bandwidth. So the research of this paper based on DSP, various techniques and methods used to optimize AVS decoder are presented and provided for the further hardware implementation. In this paper, the research of AVS video is based on the open source codec provided by Union Source Company. We implement and optimize the AVS video decoder on a generational multimedia programming and fixed-point processor TMS320DM642 provided by TI company.The paper describes the principle and key technologies of AVS coding standard, and introduces the scheme of the AVS decoder and its porting to DSP. Then analyses the complexity of AVS decoder and finds out the the critical modules are motion compensation, integer transform, quantization and loop filter. In addition to various optimization methods which are used to implement the decoder on DSP platform are presented, many techniques focused on these modules to optimize the decoder are also discussed in detail, such as using intrinsics, algorithms, code structures and memory optimization methods. The experiment results show that our design improves the performance efficiency, and the decoder is significantly speeded up after optimizing AVS decoder. And finally the optimized decoder can realize the real-time decoding (Jizhun profile) on DM642 platform.