任意视点电视系统中深度数据的自适应非均匀量化方法  

Adaptive Non-Uniform Quantization of Depth Data in Free-Viewpoint Television System

作　　者：刘晓仙[1] 常义林[1] 杨海涛[1] 元辉[1] 

机构地区：[1]西安电子科技大学综合业务网理论及关键技术国家重点实验室,西安710071

出　　处：《西安交通大学学报》2010年第2期82-87,共6页Journal of Xi'an Jiaotong University

基　　金：国家自然科学基金资助项目(60772134);高等学校学科创新引智计划资助项目(B08038)

摘　　要：为了降低现有任意视点电视系统中深度数据量化所带来的量化失真,提出一种自适应非均匀深度量化方法.首先根据深度数据的统计特性,选取有效深度子空间;然后对有效深度子空间重新进行精细划分,得到对应的自适应量化表;最后使用该自适应量化表对原深度数据进行自适应量化.实验结果表明,与现有的均匀量化和普通非均匀量化方法相比,使用所提出的方法可使深度数据的量化信噪比平均提高20 dB.因此,自适应非均匀量化方法能够显著降低深度数据量化误差,有助于在任意视点电视系统的接收端获得高精度深度数据,从而得到主观效果更好的虚拟视图.An adaptive non-uniform quantization method is proposed to reduce the distortion due to the depth data quantization in free-viewpoint television （FTV） system using the ordinary nonuniform quantization method. The proposed method determines the valid depth sub-spaces according to the statistical distribution characteristics of the depth data. Then, these depth subspaces are finely divided to form an adaptive quantization look-table. Finally, the original depth data are adaptively quantized using the quantization look-table. Experimental results and comparisons with the ordinary non-uniform depth data quantization method show that the proposed method improves the quantization signal-to-noise ratio of the depth data by an average of 20 dB. Thus the adaptive non-uniform quantization method can significantly reduce the quantization distortion of the depth data, and benefits the high-accurate reconstruction of the depth data and the Better subject visual effect of the synthesized image at the receiver of the FTV system.

关 键 词：任意视点电视 深度数据 深度子空间 自适应非均匀量化 

分 类 号：TN919.8[电子电信—通信与信息系统]