针对新一代高效视频编码标准(high efficient video coding,HEVC)逐一划分、逐层对预测模式进行RDO过程计算复杂度高的问题,提出基于结构张量和活动值的HEVC-SCC帧内快速算法.首先利用屏幕内容图像中均匀和小的全局运动区域常用大尺寸单元CU编码,复杂或大的全局运动区域常用小尺寸CU编码的特点,通过提取能够表示CU均匀性的结构张量,研究结构张量与CTU深度划分的联系,在CTU进行遍历不同深度下的编码模式前先对当前深度CU计算结构张量值,通过结构张量值判断是否跳过当前深度下的遍历率失真优化(RDO)的过程.其次利用屏幕内容和自然内容图像纹理特性不同,屏幕内容常含有水平或垂直的边,提出了基于图像活动值的屏幕内容帧内编码模式决策.通过计算图像的编码单元(coding unit,CU)的水平活动值、垂直活动值,对CU进行判别以跳过遍历所有预测模式的过程.所提的算法经过实验测试,在全帧内(all intra)配置下,与SCM-8.8算法相比能减少26.65%的编码时间,而BDBR仅增加1.95%.
A Fast Intra-Frame Algorithm for HEVC-SCC Based on Tensor Structure and Activity Value
Aiming at the problem that HEVC-SCC is divided one by one and RDO calculation complexity of prediction mode layer by layer is high,a fast intra-frame algorithm for HEVC-SCC based on tensor structure and activity value is proposed.Firstly,the uniform and small global motion regions in the screen content images are commonly encoded by large-size CU units,and the complex or large global motion regions are commonly encoded by small-size CU units.By extracting the structural tensor that can represent CU uniformity,the re-lationship between the structural tensor and the depth division of CTU is studied,and the appropriate threshold is obtained by analyzing the structural tensor of CU at different depths,at the current depth is judged.Secondly,according to the different texture characteristics of screen content and natural content images,the screen content often contains horizontal or vertical edges,and the intra-frame coding mode decision of screen content based on image activity value is proposed.By calculating the horizontal and vertical activity values of the cod-ing unit(CU)of the image,the CU is distinguished.Under the all intra configuration,the proposed algorithm can reduce the encoding time by 26.65%compared with the SCM-8.8 algorithm,while the BDBR only increases by 1.95%.
high efficiency video codingvideo codingtensor structureactivity valuea fast intra-frame algorithm
NETL
NSTL
万方数据
