基于深度的虚拟视失真估计方法  

Depth Based Virtual View Distortion Estimation Method

作　　者：李春华[1,2] 安平[1,3] 左一帆[1] 尤志翔[1] 张兆扬[1,3] 

机构地区：[1]上海大学通信与信息工程学院,上海200072 [2]河北科技大学信息科学与工程学院,河北石家庄050018 [3]新型显示技术及应用集成教育部重点实验室,上海200072

出　　处：《电视技术》2014年第11期215-218,223,共5页Video Engineering

基　　金：国家自然科学基金重点项目(60832003);国家自然科学基金项目(61172096);上海市科委项目(12DZ2293500)

摘　　要：提出一种基于深度的虚拟视失真估计方法,将深度图宏块划分为平坦宏块和非平坦宏块,结合各自特点分别用频域块级方法和时域像素级方法估计由深度失真引起的虚拟视失真。实验证明,该方法估计准确度较高,且复杂度比像素级模型有所降低。它对于提高深度视频编码效率和优化多视+深度(MVD)码率分配具有重要指导意义。A depth based virtual view distortion estimation method is proposed in this paper. The coding macro-blocks in depth map are classified into even macro-blocks and uneven macro-blocks. The block-level frequency domain distortion estimation method is used in even macro-blocks while pixel-level time domain method in uneven macro-blocks. The advantages of the two estimation methods are combined effectively. The simulation experiments confirme that the proposed method can estimate the virtual view distortion caused by depth map distortion with good balance between calculate accurate and compute complexity. It provides importance to improve 3DV code efficiency and rate allocation performance.

关 键 词：三维视频 自由视点电视 失真估计 虚拟视失真 深度图失真 

分 类 号：TN94[电子电信—信号与信息处理]