Research On Optimization Of Merge Tool In VVC

On the basis of High Efficiency Video Coding(HEVC),a large number of new technologies have been introduced and improved in Versatile Video Coding(VVC).By obtaining the motion vector of the current coding block through the surrounding coded blocks,the Merge tool can significantly improves the efficiency of inter prediction.However,there still existe many problems in Merge tools.Among them,the Combined Inter and Intra Prediction(CIIP)mode has low prediction efficiency for complex texture areas,and Affine Motion Compensated Prediction(AMC)mode has high computational complexity.This thesis conducts research on the above problems,and the main contents and innovations are as follows:(1)In order to improve the prediction performance of CIIP mode in complex texture areas,an improved CIIP mode combined with Geometric Partitioning Mode(GPM)is proposed.By introducing the flexible division method and efficient representation mechanism of GPM,the current coding block is divided into two regions for intra and inter prediction respectively.For the intra prediction region,an improved Most Probable Mode(MPM)list is designed to improve the intra prediction performance.For the inter prediction region,the original CIIP mode is used to obtain the predicted value.In addition,an integer fusion matrix,which fuses the intra prediction value and the inter prediction value to obtain the final CIIP mode prediction value is designed.The experimental results show that compared with the VTM-8.0reference model,the BD-rate of the Y component of the proposed algorithm is reduced by 0.12% on average,and the BD-rate of the U and V components is nearly unchanged,and the overall encoding time is only slightly increased.(2)In order to reduce the computational complexity of AMC moo de,a fast decision algorithm based on spatial correlation is proposed.Through the analysis of AMC mode,it is found that the decision of AMC is correlated to the motion characteristics of the current coding block.Based on high spatial correlation between the current coding block and adjacent blocks,it is proposed to use the depth of the current block and the prediction mode of adjacent blocks,and regional dispersion to judge the motion characteristics of the current block,and to decide whether the current block can skip the AMC mode.In this way,the computational complexity of the inter prediction process can be reduced,and the encoding time can be saved.The experimental results show that compared with the VTM-8.0 reference model,the overall coding time of the proposed algorithm has been decreased by 10.29%,with the BD-rate of the Y,U and V component only increasing by 0.11%,0.10%,and 0.08% on average respectively.