Research On Efficient Intra Coding Scheme Based On H.266/VVC

As the latest generation of video coding standards,H.266/ Versatile Video Coding(VVC)adopts the Quad Tree with nested Multi-type Tree(QTMT)division structure,and the number of intra modes are expanded from 35 to 67.These encoding tools endow VVC with better compression performance,while significantly increasing the computational complexity.To further reduce computational complexity,this paper studies the Coding Unit(CU)partition decision algorithm and intra mode decision algorithm respectively.The main work is as follows:(1)Aiming at the problem of unbalanced performance of traditional image feature-based CU partition algorithm,this paper proposed a texture-based fast CU partition decision algorithm.First,some unnecessary partition types are terminated or skipped according to the partition depth of adjacent reference blocks.Then,the monotonicity of the texture is used to decide the QT partition.Finally,the optimized Canny operator is used to extract edge features in a more precise way,thus providing a basis for skipping non-promising partition directions.Experimental results show that the proposed algorithm saves coding time by 51.14% on average,while BDBR increased by 1.33% on average.(2)Aiming at the bottleneck in compression efficiency faced by fast algorithms based on traditional image features,this paper proposed a visual perception-oriented fast CU partition decision algorithm.First,the color complexity is taken as the control factor to improve the contrast masking of pixel domain-based JND model,so as to further remove the visual redundancy in a way that is more in line with the characteristics of visual perception.Then,a visual saliency model based on motion states is designed to reasonably allocate bits to the current region according to different motion states to improve subjective quality.Finally,the decision tree is constructed by combining the above model with texture features to further reduce the overall computational complexity.The experimental results show that the proposed algorithm saves coding time by 48.71% on average with 0.80%BDBR increase.(3)Aiming at the inefficiency of intra mode decision-making,this paper proposed a fast intra mode decision algorithm based on Limited-Broyden-Fletcher-Goldfarb-Shanno(L-BFGS).First,the angle modes are classified according to the direction of texture.Then,texture complexity is used to determine the direction of the main and secondary textures(if it existed)in the current region,thus significantly reducing the execution rate of Rough Mode Decision(RMD).Finally,L-BFGS adaptively searches for an angle mode with the lowest RD-cost in the texture direction to further reduce the execution rate of Rate Distortion Optimization(RDO).Experimental results show that the proposed algorithm saves coding time by 26.18% on average with 0.42% BDBR increase.