一种VVC帧内编码单元快速划分算法  

Fast Division Algorithm of VVC Intra-coding Unit

作　　者：陶浩然 路锦正[1,2] 李意弦 TAO Hao-ran;LU Jin-zheng;LI Yi-xian(School of Information Engineering,Southwest University of Science and Technology,Mianyang 621010,China;Robot Technology Used for Special Environment Key Laboratory of Sichuan Province,Mianyang 621010,China)

机构地区：[1]西南科技大学信息工程学院,四川绵阳621010 [2]特殊环境机器人技术四川省重点实验室,四川绵阳621010

出　　处：《小型微型计算机系统》2021年第7期1470-1474,共5页Journal of Chinese Computer Systems

基　　金：国家自然科学基金项目(61601382,61401379)资助;四川省科技厅支撑计划项目(2017GZ0316)资助。

摘　　要：为了降低下一代通用视频编码(VVC)帧内预测编码单元(CU)划分的计算复杂度,提出一种基于梯度幅值相似度的CU快速划分方法.首先,计算当前编码单元下层的四个子编码单元的平均梯度幅值相似度偏差(M GM SD),根据该信息来确定当前编码单元是否进行四叉树划分或不划分.其次,当不满足四叉树划分和不划分的条件时,通过遍历得到三叉树划分和二叉树划分的子块像素方差的方差,根据该信息来选择二叉树和三叉树中最佳的划分方式.在全I帧条件下,本文方法与VTM7.0(VVC Test Model 7.0)标准模型相比,编码时长平均降低了50.69%,在大幅降低编码复杂度的同时码率仅增加1.36%.In order to reduce the computational complexity of the next generation universal video encoding(VVC)in-frame prediction coding unit(CU),a CU rapid division method based on gradient amplitude similarity is proposed.First/calculate the average gradient similarity deviation(MGMSD)of the four sub-eoding units at the lower level of the current encoding unit,and use this information to determine whether the current encoding unit is four-tree or undivided.Secondly,when the conditions for the division and non-division of the four-tree are not met,the best division of the binary tree and the trident tree is selected according to this information by traversing the variance of the pixel variance of the sub-block of the Trident and the division of the trident tree.Under fullⅠ-frame conditions,the coding time of this paper is reduced by an average of 50.69 percent compared to the VTM 7.0(VVC Test Model 7.0)standard model,and the code rate increases by only 1.36 percent while significantly reducing the complexity of coding.

关 键 词：VVC 帧内预测 CU划分 平均梯度幅值相似度偏差 多类型树 

分 类 号：TP919.81[自动化与计算机技术]