Research On Optimization Algorithm Of HEVC Intra Frame Mode Decision And CU Division

The new generation video coding standard,High Efficiency Video Coding(HEVC)achieves about 50% bitrate reduction compared to the previous generation standard h.264 under the same perceptual video quality by adopting many new techniques.However,it also greatly increases the coding complexity.Therefore,how to balance the coding quality and coding complexity has become a research hotspot in the current HEVC.In this thesis,the key part of HEVC,intra prediction,is studied.The algorithm is optimized from two aspects: block partition and mode selection of intra prediction,in order to achieve the purpose of reduce the encoding complexity and encoding time.The main research results are as follows:Aiming at the high complexity of the partitioning of HEVC encoder x265 intra coding unit(CU),this thesis proposes an optimization algorithm(FCU-SSCU)based on the complex similarity of the sub-blocks of the current CU and the total number of bits and quantization parameters required for its encoding in oder to terminate CU partition early.First,the difference matrix is used to calculate the texture complexity and similarity of the sub-blocks of the current CU,Then,the CU partition is terminated in advance according to the correlation among the quantization parameter of the current coding block and the number of coded bits and the similarity between sub-blocks,reducing the calculation of the redundancy depth.The experimental results show that compared with the standard algorithm of x265,the algorithm can reduce 22.23% of the intra-frame encoding time on the premise of basically unchanged video quality.In order to solve the high complexity of intra-mode selection in HEVC,two fast algorithms are proposed in this thesis from the perspective of reducing candidate list and mode prediction.The optimization algorithm(FMS-SATD)of mode selection based on Sum of Absolute Transformed Difference(SATD)is proposed in this thesis.The algorithm is to further reduce the number of candidate lists selected by the prediction unit(PU)mode by increasing the SATD calculation,thereby reducing the rate distortion cost calculation.The experimental results show that the proposed algorithm(Combination of FCUSSCU and FMS-SATD)can reduce the average intra-frame coding time by 32.34% on the premise that the video quality is basically unchanged.The optimization algorithm based on the statistical information mode(FMS-CAI)is to used to predict the mode of the current CU according to the mode information of the adjacent CUs when the complexity and the depth are basically the same,thereby reducing the redundant mode calculation.The experimental results show that compared with the x265 standard algorithm,the intracoding time is reduced by an average of 19.93% when the performance loss is negligible.To sum up,the optimization algorithms of block partition and mode selection of HEVC intra module achieves good results.Under the premise of guaranteeing the video quality,it reduces the HEVC intra coding complexity effectively at the cost of negligible coding performance.