| In recent years,more and more virtual reality products have come out,and virtual reality technology has received extensive attention.Virtual reality 360-degree video(referred to as 360-degree video)is the main content of virtual reality technology in the field of video coding.The Joint Video Experts Team(JVET)proposes a coding framework for 360-degree video coding that uses the latest high-efficiency video coding standard(HEVC),but with 360-degree video with high resolution(4K,6K,8K),high frame rate(60fps),high bit depth(10bit)and other features,resulting in 360-degree video encoding complexity is extremely high.In addition,the HEVC standard itself has a high coding complexity.In this paper,the intra prediction process and the sample adaptive compensation process of 360-degree video are deeply studied,and the following results are obtained:(1)A fast algorithm for 360-degree video intra prediction based on the similarity between child blocks and parent block is proposed.360-degree video should be projected as a flat video before encoding,but this process can cause projection distortion,the degree of distortion depends on the latitude position of the coding tree unit.360-degree video is different from the objective video quality metrics of traditional video.WS-PSNR has dealt with the proj ection distortion and introduced the concept of sphere video weight.The similarity between child blocks and parent block can reflect whether the parent block needs to be divided by the quadtree.The similarity combined with the weight of the spheres can realize the difference processing between the polar and equatorial regions of the sphere video.The early termination of the PU is achieved by selecting an appropriate threshold.The experimental results show that the algorithm achieves an average 32%time reduction in all intra configuration,and only 0.29%BD-rate increases.(2)A fast sample adaptive offset(SAO)algorithm for 360-degree video is proposed.Based on the analysis of the complexity of SAO process,a simplified SAO process is designed.A two-stage selection process is performed according to the rate distortion cost(RD-cost)and the equirectangular projection(ERP)characteristics,and finally the CTU that performs the simplified SAO process is determined.The experimental results show that the SAO processing time is reduced by 57%,54%,65%and 64%on average,under the configurations of all intra,random access,low delay B,and low delay P,which only 0.05%,0.19%,0.42%and 0.53%BD-rate increases,respectively. |