Research On Fast Algorithm For Inter Coding Of Versatile Video Coding

With the continuous development of the video industry,video is changing people's lives in terms of social interaction,learning,and entertainment.However,the ever-increasing amount of video data has brought huge transmission pressure to the communication network.In the face of this problem,it can be solved by compressing the video and transmitting it.To this end,MPEG and VCEG jointly formulated H.266/Versatile Video Coding(VVC)to provide a video coding standard with higher compression efficiency.As the latest video coding standard,VVC has introduced a series of new technologies to further improve compression efficiency,but these new technologies have also greatly increased the coding complexity and brought challenges to actual use.In order to reduce the complexity of VVC coding,this thesis studies the VVC video coding technology and proposes two fast algorithms for inter coding.First of all,this thesis studies the coding unit(CU)partition process of VVC inter coding.During the study,it is found that because VVC adopts a more flexible CU partition structure and a deeper division depth,the coding of VVC in the block partitioning process complexity is extremely high.In response to this situation,this thesis proposes a fast algorithm for CU partition based on the spatial and temporal correlation of video sequences.By extracting the motion vector of the region corresponding to the reference frame in the time domain,analyzing the motion vector distribution characteristics of the region,and then selecting one or more division modes in advance to reduce the number of division modes that need to be traversed;analyzing the current CU adjacent to the spatial domain the consistency of the prediction mode of the determined CU of the same size,and the cost of rate distortion is measured to determine whether to enter the next depth for traversal.If there is consistency and the cost value of rate distortion is not high,then the next division depth will not be entered for calculation,reducing the division depth of the traversal.The experimental results show that under the configuration of random access mode,the algorithm in this thesis reduces the peak signal-to-noise ratio by 0.05d B and the coding output bit rate increases by 1.67%compared with the original algorithm,saving 37.62%of the coding time.Secondly,this thesis studies the motion estimation process of VVC inter-frame coding.In the research,it is found that because VVC adds four-parameter and six-parameter affine motion estimation,the coding complexity of the motion estimation process is greatly improved.In response to this situation,this thesis proposes a fast algorithm for motion estimation based on the parent CU prediction mode and the best prediction direction of the current CU.Determine whether to skip four-parameter and six-parameter affine motion estimation by referring to whether the parent CU prediction mode of the current CU is skip mode;when performing six-parameter affine motion estimation,refer to the best prediction direction of four-parameter affine motion estimation,then decide whether to skip part of the prediction direction of the six-parameter affine motion estimation.The experimental results show that under the random access mode configuration,the algorithm in this paper can save 6.25% of the coding time,while the coding output bit rate is increased by 0.08%,and the peak signal-to-noise ratio is reduced by 0.01 d B.