AVS Software Decoder Optimization Based On XSBase270

AVS is the audio video coding standard made with fully patent by the Audio Video Coding Standard Workgroup of China in 2003.The development of China's multimedia industry has been restricted by the patents of H.264,MPEG-2.This problem can now be solved with AVS's patent pool policy.AVS,which has occupied an important position in many fields in China,such as digital TV,IPTV, multimedia communication,video surveillance,is really meaningful to the future of China's multimedia industry.Recent researches on AVS products,AVS key technology are presented in this dissertation and so does the comparison between AVS and H.264,MPEG-2.Based on rm-52j refer decoder,the AVS decoder on XSBase270 is optimized on algorithm and by assembly language.AVS decoder is first optimized on algorithm.During luma interpolation,picture padding is used in order not to clip the pixel coordinates.The implementation of chroma sub-pixel interpolation is simplified by avoid interpolation when vertical or horizontal motion vector is integer.Also branch is implemented with table,which reduces the consuming time of branch.Data exchange between CPU and SDRAM cost more time on embedded system than on PC.In this dissertation,The process of intra prediction and reconstruction is combined to avoid data store after ICT and data read before reconstruction.WMMX is a set of SIMD instruct suitable for large amounts of data process.Interpolation and ICT functions are realized with WMMX instructs and so get higher running speeds. With all the optimizing methods mentioned above,the QCIF decoding rate of AVS decoder on XSBase270 can reach 24 fps or above which means real-time decoding.