Fast Algorithm For RDOQ In AVS2 Video Coding And Implementation Of Frame-level Parallel Decoding Based On OpenMP

AVS2(Audio Video Standard 2)is China's second-generation video coding standard with independent intellectual property rights,which has comparable performance in comparison with the latest international standard HEVC(High Efficiency Video Coding).AVS2 doubles performance over previous generation H.264 or AVS1,and it has special optimization for monitoring the scene and 3D video coding.Because of the high computational complexity,AVS2 cannot meet the need of real-time.Therefore,the study of both encoding and decoding is of great significance.Firstly,the thesis introduces the history and the latest development of video coding standards.the basic principle and key technology of AVS2 standard are introduced in detail,and the implementation process of important modules is analyzed with official source code RD14.0.The AVS2 standard is a lossy video compression technology,the main reason for video quality loss is quantization module.Meanwhile it also affects the encoding bitrate.Two quantification methods are used in RD14.0.One is rate-distortion optimization quantization(RDOQ),which firstly calculates multiple candidate quantized value for each pixel value,and then combines the rate-distortion optimization technique to preprocess each candidate value to estimate the code rate and distortion,finally choose the best quantized value.Although the RDOQ process greatly improves the video compression rate,the high computational complexity also reduces the encoding efficiency.The other quantization is traditional scalar quantization,which has low complexity and compression ratio.The thesis compares and analyzes relations between two kinds of algorithm,and proposes three optimized schemes based on rate-distortion optimization quantification algorithm,at last implements it in code RD14.0.The test results show that both schemes can effectively improve the coding speed and maintain a high compression ratio.However,different schemes have different advantages and disadvantages.Experimental results show that the code speed of the new improved scheme is increased by an average of 14% to 21%,and the peak signal-to-noise ratio is reduced by 0.022 dB to 0.058 dB on average.The above analysis is based on the AVS2 encoder.In addition to that,the thesis also parallel optimization for the AVS2 decoder.The AVS2 does not support wavefront parallel processing or Tile-level parallelism as the HEVC standard does.The thesis first analyzes the avs stream structure and the decoding process,and uses the OpenMP interface to modify the decoder program to achieve frame-level parallel decoding of the avs stream in the All_intra encoding configuration.Finally,the experimental results show that under four threadsdecode at the same time on a 4-core CPU platform,the parallel speedup can reach 2.32 to3.07.