| The new generation video coding standard H.266/VVC has been promulgated in 2020.Compared with the previous generation video coding standard H.265/HEVC,its compression performance has been improved by more than 40%.Its excellent performance and friendly support for ultra-high definition video,screen video and panoramic video depend on the introduction of a large number of new coding tools.However,the performance advantage comes with high computational complexity.The high-efficiency distortion optimization scheme can ensure the compression performance of the encoder and effectively reduce the computational complexity.Intra prediction is the key module of video coding,and its computational complexity accounts for more than 30% of the whole encoder.Therefore,this paper studies the optimization of intra prediction module of video coding standard H.266/VVC.The specific research contents are as follows:Based on the traditional statistical law analysis method,this paper puts forward an optimization scheme for the modules of CU division,intra mode selection,luma prediction and chroma prediction in intra coding.The specific algorithms include: the CU division optimization algorithm based on image gradient,which optimizes the multi type tree direction division of CU by establishing the relationship model between image gradient and CU division type;The mode selection optimization algorithm based on encoded information,including intra and inter mode selection optimization algorithm based on parent block information and intra mode selection optimization algorithm based on peer block information,mainly optimizes the mode selection process of non-I frame CU and CU with peer block;The luma prediction optimization algorithm based on rough selection step size mainly optimizes the angle prediction rough selection in brightness prediction by adjusting the rough selection step size;The intra coding optimization algorithm of chroma prediction based on rough selection list mainly optimizes the fine selection of chroma prediction by reducing the rough selection list.The optimization algorithm based on statistical law analysis proposed in this paper can reduce the time complexity of 31.02% under the performance loss of 1.54% compared with the slow gear under random access condition.Based on the extreme forest tool,this paper proposes an optimization algorithm for the CU division process of intra coding.Firstly,a limit forest model is established to judge the division type of CU.The division types are divided into four types: no division,quad division,multi fork horizontal division and multi fork vertical division.The division methods with low probability are skipped according to the output results of the model;Then,the second limit forest model is established to judge whether to choose binary partition or trigeminal partition when CU is divided into multi tree.The model can be used after the first model to further reduce the number of candidate partition types of CU.The experimental results show that the optimization algorithm based on limit forest proposed in this paper can reduce the time complexity of 26.02% under the performance loss of 0.49% compared with the slow gear of VVen C1.0.0 encoder under all intra condition;Compared with the medium gear of VVen C1.0.0 encoder under all intra condition,it can reduce the time complexity of 13.99%under the performance loss of 0.42%. |
