基于复杂度和空域相关性的虚拟现实视频CU划分快速算法  

Fast CU Partitioning Algorithm for Virtual Reality Video Based on Complexity and Spatial Correlation

作　　者：刘志[1] 杨柯[1] 付显亚 LIU Zhi;YANG Ke;FU Xianya(Col.of Information,North China Univ.of Tech.,100144,Beijing,China;Yi Xinda(Beijing)Technology Co.,Ltd.,100020,Beijing,China)

机构地区：[1]北方工业大学信息学院,北京100144 [2]宜信达(北京)科技有限公司,北京100020

出　　处：《北方工业大学学报》2020年第2期45-51,共7页Journal of North China University of Technology

基　　金：北京市自然科学基金项目“面向交互的虚拟现实视频编码与传输关键技术研究”(4202018).

摘　　要：在等距柱状投影格式下,虚拟现实360度视频的两极区域会造成过度拉伸,引入大量冗余数据,影响编码速度.论文分析了等距柱状投影下虚拟现实360度视频图像特点,设计了一种度量编码树单元复杂度的新指标,利用复杂度指标和视频空域相关性预测编码树单元深度,跳过不必要的遍历范围,实现编码单元的快速划分.实验结果表明,与最新的HM16.20参考模型相比,所提出的算法可以减少37.25%的编码时间,而BD-rate仅增加0.74%,视频图像质量几乎没有损失.In the equirectangular projection(ERP) format, the polar regions of virtual reality 360-degree video are overstretched, which result in introducing a lot of redundant data and reducing the encoding speed. In the paper, the characteristics of video images in the ERP projection format are studied and a new metric is designed to measure the complexity of coding tree unit(CTU). Based on the complexity metric and the spatial correlation of video image, the depth of a CTU is predicted to skip the unnecessary traversal search during coding unit(CU) partition so as to speed up the partition process. Experimental results demonstrate that, compared with the latest reference model HM 16.20, the proposed algorithm can reduce the encoding time by 37.25%, while the BD-rate only increases by 0.74%, and video image quality is not impaired.

关 键 词：虚拟现实 360度视频编码 CU快速划分 CTU复杂度 空域相关性 

分 类 号：TN919.8[电子电信—通信与信息系统]