Research On 360 Degree Video Coding Algorithm Optimization Based On CTU Complexity

Virtual reality 360-degree video requires ultra-high resolution to provide realistic feeling and dynamic perspective.The huge amount of data presents challenges to coding and transmission.In the ERP projection format,the 360-degree video image is stretched to varying degrees,generating a lot of redundant data,which significantly affects encoding speed and encoding performance.Therefore,researches on coding optimization of 360-degree video have received widespread attention.In order to improve coding performance and reduce coding time,a 360-degree video coding optimization algorithm based on CTU complexity is proposed to optimize the quantization process and the CU partition process in HEVC.The main research contents and innovations are as follows:First,for the problem of QP offset selection adaptability during the quantization process,the characteristics of video images in the ERP projection format are researched,a new index for measuring CTU complexity is designed,an adaptive QP offset selection algorithm based on CTU complexity is proposed.In this algorithm,CTUs are classified into five levels according to complexity,and different QP offset values are determined for each level.In this way,the flat blocks in the equatorial region and the complex blocks in the polar region can be quantified effectively.The coding performance is enhanced by improving the quality of the complex region and reducing the bit rate of the flat region.Experimental results demonstrate that,compared with the latest reference software HM 16.20,the proposed algorithm can achieve 0.40 d B WS-PSNR improving and 1.99% BD-rate saving.Second,for the problem of high complexity caused by traversing all the depth during the CU partition process,the correlation between coding blocks and adjacent blocks at different levels is studied,the proportion of four partition depths at different levels is counted,and a fast CU partition algorithm based on CTU complexity and spatial correlation is proposed.In this algorithm,different depth prediction methods are determined according to the level of CTU.Level 1 traverses 2 and 3 depths,level 5 traverses 0 and 1 depths.The corresponding reference CTUs are selected to calculate the average depth for level 2,3,4,the predicted depth is obtained according to the threshold interval.In this way,the depth traversal process is simplified and the coding complexity is reduced.Experimental results demonstrate that the proposed algorithm can obtain 37.25% coding time saving.