Research On AVS2 Efficient Intra Prediction Mode Selection Algorithm And Arithmetic Coding Optimization

AVS2(Audio Video Standard II)is a new generation of digital Audio and Video coding and decoding Standard released after AVS1/AVS+ series,among which AVS2-P2,namely "efficient multimedia coding for information technology",is part of Video.Its performance is basically equivalent to the h.265/HEVC standard in the same period,but compared with the previous generation standard AVS1,its coding compression efficiency is more than double.As a compression standard for video with high definition,ultra high definition and high dynamic range,it promotes the development of China's information industry.AVS2 achieves better compression efficiency by using more advanced coding techniques,such as more complex block partitioning,more prediction modes,etc.,which greatly increases the coding complexity.The coding efficiency of the reference code of the new standard RD14.0 is far from the requirement of real-time coding,and it is necessary to optimize the coding algorithm.In this paper,the current research results at home and abroad is analyzed,and the key coding technology of AVS2 standard is briefly introduced.Combined with the whole coding process in RD14.0 reference software,the AVS2 intra prediction mode selection algorithm is introduced in detail.The coding complexity distribution of the intra prediction is analyzed,and the intra prediction mode selection algorithm based on the direction texture complexity(Algorithm I)is proposed.The algorithm reduces coding complexity by reducing the set of candidate modes of the RMD and RDO processes.In the RMD process,33 kinds of prediction modes were divided into five groups according to the direction texture complexity of CU,and the number of traversal mode of the RMD is reduced by a strategy of selecting one of the five groups.In the RDO process,combined with MPMs and mode hit ratios,the number of prediction mode rate distortion optimizations is reduced.On the basis of RD14.0,the algorithm saves about 41% of the coding time on average,and the PSNR of the luminance only drops by about 0.06 dB,and the loss of coding quality is small.Based on the analysis of the shortcomings of the above algorithm,the judgment of the texture direction of large size CU is prone to a large deviation,which greatly increases the BD-rate of some video sequences.Thus,an intra prediction mode selection algorithm based on mode distribution(algorithm II)is proposed.An intra prediction mode selection algorithm based on mode distribution(algorithm II)is proposed.According to the pattern utilization distribution,33 kinds of prediction models were divided into two groups of common modes and three groups of unusual modes.Abandoning the original selection strategy of the RMD process and RDO process,the RMD screening is performed in advance for each intermediate mode of the five groups of modes,and then the mode group that needs to perform the second RMD process is determined,and the RDO process is similar to the algorithm I.The coding time of the algorithm is saved by about 44%,and the evaluation of PSNR and BD-rate is obviously better than algorithm I.Finally,the CABAC algorithm of the entropy coding in the lowest layer of the coding framework is analyzed.Since it is called at a high frequency,there is a positive effect on optimizing this part.Every symbol coding carries out the probability update process in the CABAC algorithm,which results in high complexity.Therefore,the CABAC probability update acceleration algorithm is proposed which adaptively adjusts the number of updates of probability.The algorithm reduced about 8% of the encoding time of the entropy coding module of AVS2,and has less influence on the coding quality.