Optimization of AVS plus Software Decoder for High-definition Video

Advance Video Coding System Plus (AVS+), the new generation of AVS video coding standard in China, is designed to achieve significant coding efficiency improvements relative to the preceding H. 264/AVC and AVS1 standard. Although its coding performance is improved, the complexity of decoding is also increased. In this paper, we propose a parallel optimization method based on different levels of granularity to implement the optimization of AVS+ software decoder for real-time HD (1920x1080) video decoding. Firstly, for modules with computational intensity in the AVS+ decoder, single-instruction-multiple-data (SIMD) instructions are exploited to optimize the computational performance of these modules in way of data-level parallelism. Secondly, by decomposing the decoding process into several parts that can be executed in mode of parallel pipeline, multi-thread optimization is used to achieve a higher level parallelism, thus further improving the decoding performance. The experimental result shows that the optimized decoder reaches an average decoding frame rate of 38 fps for high-definition (1080P) videos, and gives a total speed-up of more than 170% compared with the original AVS+ decoder.