Avs Video Coding Algorithm In Parallel To Achieve

The new generation video coding standard AVS(Audio and Video coding Standard) which was proposed by "Digital Audio Video Coding Standard Working Group", is being more and more extensively applied and researched in the filed of industry and academic because of the advantages such as: high performance, low computational complexity, and low licensing cost, etc. Using the latest video coding technology, the encoding speed of AVS is much lower than the previous generation video coding standard. Now, with the popularity of multi-core processors, multi-threaded parallel programs will effectively improve the encoding speed. OpenMP is a multi-thread programming standard for shared memory multi-processor system, and provides a simple and efficient multi-threaded application programming interface for programmer.To improve the AVS encoding speed, parallel computing theory, AVS coding standard and OpenMP multi-threaded programming standard are described, and three tasks have been completed:Firstly, using OpenMP, the parallel encoding algorithms including GOP-level, slice-level, frame-level are researched and realized, based on the reference model RM62f of AVS encoder. These parallel encoding algorithms have been tested on Intel quad-core processor platform, and given the relationship between the number of threads and speedup, PSNR and bit rate. Experimental results show that, parallel algorithms can improve encoding speed signifiantly compared with original serial algorithm, but they still have their own shortcomings.Secondly, a frame-level encoding algorithm based on task queuing model is presented to overcome the frame-level's shortcomings in scalability and load balancing. Experimental results show that, the speedup of this algorithm can obtain 3.9x on Intel quad-core processor platform. This parallel encoding algorithm can improve encoding speed and CPU utilization greatly, at the same time SNR and bit rate remain completely unchanged.Finally, in order to further improve the encoding speed, this thesis realizes SIMD parallel optimization of AVS video coding, using Intel multimedia instruction set SSE2 and SSE4. Experimental results on Intel quad-core processor platform show that, after Intel multimedia instruction set optimization, AVS video encoding speed increase signifiantly. Especially, SSE4 optimization can make the speedup up to 14.8x, combined with the frame-level encoding algorithm based on task queuing model.