Research On Intra Coding Fast Algorithm Of H.266/VVC

Video applications such as video broadcasts,video on demand,and video conferencing has become commonplace and play an indispensable role in life.However,the compression performance of the High Efficiency Video Coding(HEVC)standard is hard-pressed to meet market application demand with people’s pursuit of resolution,frame rate and color saturation.To address this problem,the Joint Video Exploration Group(JVET)developed the Versatile Video Coding(VVC)standard to reduce the pressure on video transmission and storage.On the basis of HEVC,many new coding technologies or improves the original algorithms are adopted by VVC.The improve of compression performance is paralleled by an increase in computational complexity,which hinders the application and industrialization of VVC.In order to reduce the computational complexity of VVC,the thesis explores the fast intra prediction coding techniques for VVC,and two fast algorithms are proposed.Firstly,a fast CU split algorithm based on image texture characteristics and residual coefficient distribution is proposed.The algorithm judged the texture of the CU by the absolute average difference.If the current CU texture is simple,the depth traversal for the CU is terminated;If the current CU texture is rich,the current CU would be roughed divided into upper-lower partition and left-right partition.Then,the similarity of different partitions is calculated and compared based on the residual coefficient distribution of the current CU,and the horizontal partitioning or the vertical partitioning is skipped according to the result of the similarity.Finally,the performance of proposed fast algorithm is tested under common test conditions recommended by JVET.Experimental results demonstrate that if the coding structure is configured as all intra,the proposed fast algorithm achieves 52.1% of the coding time saving on average,while the loss of PSNR is only 0.07 d B,and the BDBR is 1.41%.Secondly,a fast intra prediction mode decision algorithm based on multiple reference lines and statistical characteristics is proposed.The algorithm roughly determines the optimal intra prediction mode based on the reference samples line.If the value of reference sample lines compliance with uniformly distributed,the directional prediction mode would be excluded;If the value of reference sample lines not compliance with uniformly distributed,the RD list is optimized and the number of candidate modes is reduced,based on the relationship between the MPM list and the optimal intra prediction mode.Finally,the performance of proposed fast algorithm is tested under common test conditions recommended by JVET.Experimental results demonstrate that if the coding structure is configured as all intra,the proposed fast algorithm achieves 32.79% of the coding time saving on average,while the loss of PSNR is only 0.16 d B,and the BDBR is 1.25%.