Research And Realization Of H.264/AVC Software Decoder Optimization

The Joint Video Team (JVT) which consists of members of the Moving Picture Experts Group and the Video Coding Experts Group (MPEG and VCEG) have developed a new video compression standard H.264/AVC. H.264/AVC provides better compression of video images, different features and more flexibility compared with existing standards such as H.261, H.263, MPEG-2 and MPEG-4. The improvement in performance leads to higher computational complexity in the new standard. Both encoder complexity and decoder complexity of H.264 are higher than MPEG-4. It is necessary to optimize H.264 codec to improve its running speed.In this dissertation, we introduced the H.264/AVC standard and its realization models and analyzed the optimization methods for software. Then the complexity of key modules in H.264 decoder is analyzed and discussed. In order to reduce computational complexity, we adjusted the flow of the decoder program, optimized modules such as macroblock decoding, CAVLC etc. for function optimization firstly; secondly, we used SIMD technique to optimize sub-pixel interpolation, integral transform for operation optimization, and carried out decoder's automatic vectorization with the help of Intel C++ Compiler.In the end, the dissertation compared the performance of the optimized decoder with the original. The results indicate the optimized decoder has increased its decoding speed (fps) about 9 after function optimization; with the help of Intel C++ Compiler, the decoding speed (fps) can increase about 35 after using SIMD technique and automatic vectorization. The optimized decoder can meet the requirement of real time applications.