The Optimization of AVS Video Decoder on the PC Platform and Improvement of Field Decoding

AVS(Audio Video Coding Standard) is developed by the Audio Video Coding Standard Working Group of China. It is the first China’s national audio video compression standard with independent intellectual property rights applied in the field of significant information industry such as high-resolution digital broadcast, high-density DVD players, internet stream media, and wireless multimedia communication. The purpose of this paper is to optimize AVS codec, thereby promoting the practical application of it.This dissertation first illustrates the codec principles of AVS, then it describes the key technologies of AVS such as intra prediction, inter prediction, transforming, entropy coding, filtering and so on. It also analyzes AVS decoder’s structure and flow. At the same time, it shows the testing results of the decoder when decoding the coded video sequences and points out the time-consuming modules.In this dissertation, we analyse the time-consuming modules carefully, then adopte various ways to optimize it. In decoder: firstly, we adjuste the flow of the decode program to reduce the times of memory access; secondly, the decoder is optimized from algorithm level, program level and SSE2 level; thirdly, in multi-core processors , we optimize the decoder through decoding in parallel.On algorithm level, Optimization methods in golomb decoder and motion compensation is proposed is proposed and implemented based on general processor; On program level, we optimize the decoder through look-up table optimization, the optimization of cycle and so on; For SSE2 level, we refine the program structures and use SSE2 instructions to rewrite the 8×8 inverse transformation, interpolation, filtering and reconstruction module to improve speed. Decoder in parallel refers to choose macroblock level, slice level and frame level parallel way to realize a multi-thread AVS decoder.After the above jobs, the decoder’s speed reaches more than 25 frames per second for the sequence of high-definition, achieveing the goal of the optimization.The paper also carries out a study on filed decoder, and brings out a new algorithm to improve it.