基于场景切换的x265帧类型判决优化

Frame Type Decision Optimization of x265 Based on Scene Change

随着短视频等视频应用的流行,视频序列常存在复杂的场景变化。现有的x265算法中,基于场景切换的帧类型判决算法在频繁场景切换、渐变场景等复杂场景下效果不佳。对x265基于场景切换的帧类型判决算法进行优化,改进了连续场景切换点的选择策略,同时提出场景持续比指标(SDR)判断是否插入关键帧,并优化了渐暗场景等特殊场景下的决策。实验结果表明优化方法相比现有方法效率有较大提升,该方法与现有算法相比,在medium和fast模式下,Y、U、V三通道在BD-Rate方面平均能得到0.531%、0.507%和0.418%,以及0.943%、0.979%和1.544%的增益。

With the popularity of short video and other video applications,video sequences often have complex scene changes.Among the existing x265 algorithms,the frame type determination algorithm based on scene change is not effective in complex scenes such as continuous multiple scenes and gradient scenes.In this paper,x265 frame type decision algorithm based on scene change is optimized,and the selection strategy of continuous scene change is improved.At the same time,the scene duration ratio(SDR)is proposed to judge whether to insert key frame,and the decision of special scene such as fading scene is optimized.The experimental results show that the efficiency of the optimized method is greatly improved compared with the existing methods.Compared with the existing algorithms,in Y,U and V channels,the average BD-rate of can get gains of 0.531%,0.507%and 0.418%in medium mode,and gains of 0.943%,0.979%and 1.544%in fast mode。