The Optimization And Implementation Of Low-complexity Codecs For High Efficiency Video Coding

With many advanced coding technologies, the compression performance of High EfficiencyVideo Coding (HEVC) is more efficient than previous H.264/AVC standard. In comparison with theH.264/AVC, the HEVC has the same peak signal noise ratio (PSNR) while its coding bitrate cansave40%for the objective quality evaluation; furthermore, the HEVC has the same visual qualitywhile its coding bitrate can save more than50%for the subjective quality evaluation. The testingresults have shown that HEVC codec achieves high compression ratio at cost of tremendouscomputational complexity. Therefore, the high-complexity technologies in HEVC leave moreselection criteria and optimization potential. The coding parameters should be reasonably andeffectively selected to improve the compression performance of HEVC codec. With the formulationof HEVC standard, the optimization of its key modules has received more attentions in bothacademia and industry. How to keep the compression efficiency at a low-complexity cost willdetermine whether HEVC codecs can be successfully applied in industry, and the HEVCoptimization is also a hot issue in academia.Now, the optimization and implementation of HEVC codec mainly focuses on the followingareas: the rate control, the low-complexity coding, the subjective quality evaluation. In this paper,the main research direction is the low-complexity codec optimization to ensure theencoding/decoding performance and reduce the encoding/decoding time at the same time. Firstly,this paper proposes a new intra prediction method based on the Sum of Absolute TransformedDifferences (SATD) criteria to speed up the intra-prediction mode decision, and thus effectivelyreduce the encoding complexity. Secondly, according to the encoding statistics of motion vectors,this paper also proposes the LD-Main integer-pixel motion estimation based on confidence intervals to reduce the calculation cost of unnecessary candidate modes. Finally, this paper also analyzes thedeblocking filter of HEVC standard and proposes the multi-core parallel deblocking filter schemebased on the directed acyclic graph set and Compute Unified Device Architecture (CUDA), and theexperiment results show that the proposed scheme can significantly improve the reconstructionspeed while the video quality has nearly no loss.