三维视频的深度图快速编码算法  

Fast Encoding Algorithm for Depth Map in Three-Dimensional Video

作　　者：韩雪[1] 冯桂[1] HAN Xue;FENG Gui(Department of Information Science and Engineering,Huaqiao University, Xiamen Fujian 361021,China)

机构地区：[1]华侨大学信息科学与工程学院,福建厦门361021

出　　处：《通信技术》2018年第3期575-582,共8页Communications Technology

基　　金：福建省自然科学基金(No.2016J01306);华侨大学研究生科研创新能力培养计划资助项目(No.1611301035)~~

摘　　要：为了解决3D-HEVC编码标准中深度图计算复杂度过高的问题,利用编码单元(Coding Unit,CU)的纹理复杂度和多视点间的相关性,对深度图编码做优化。具体地,使用Otsu’s算子计算当前CU的最大类间方差值,提前判决当前CU是否平坦,对平坦CU提前决策最佳分割尺寸;利用视点间的冗余性,参考已编码的独立视点的CU信息,对非独立视点的CU编码过程做优化。实验结果表明,提出的算法与原始3D-HEVC编码算法相比,编码时间减少25.68%,合成视点的失真可忽略不计;提出的算法与其他同类型算法相比,能够更精确区分平坦的CU和纹理复杂度高的CU,有效降低了深度图的编码复杂度。In order to solve the problem of high computation complexity of depth map in3D-HEVC coding standard,the depth map coding is optimized by using the texture complexity of Coding Unit(CU)and the correlation of between multi viewpoints.Specifically,the Otsu's operator is used to calculate the maximum inter-class variance of the current CU,so as to decide whether the current CU is smooth or not,and for the smooth CU,the optimal partition size is determined in made advance.By referring to the CU information of the coded independent viewpoint and by making use of the redundancy of between the viewpoints,the CU coding process of the non-independent viewpoint is optimized.The experimental results indicate that the proposed algorithm could reduce the coding time by25.68%as compared with the original3D-HEVC coding algorithm,and the distortion of the synthetic viewpoint is negligible.In addition,compared with other similar algorithms,the proposed algorithm could more accurately distinguish smooth CU from CU with high texture complexity,thus effectively reducing the coding complexity of depth map.

关 键 词：3D-HEVC 深度图 四叉树分割 视点间相关性 

分 类 号：TN919.81[电子电信—通信与信息系统]