一种新的深度视频处理方法  

A Novel Depth Video Processing Algorithm

作　　者：毕伟[1] 彭宗举[1] 

机构地区：[1]宁波大学信息科学与工程学院,浙江宁波315211

出　　处：《宁波大学学报（理工版）》2013年第2期33-37,共5页Journal of Ningbo University：Natural Science and Engineering Edition

基　　金：国家自然科学基金(61171163);宁波市自然科学基金(2012A610041)

摘　　要：深度视频的不连续性会影响其编码效率以及虚拟视点的绘制质量.为增强深度视频的连续性,提出了一种新的基于时空变换的深度视频处理方法：首先检测出深度视频中的运动对象;其次将彩色视频和对应的深度视频进行时空变换,对运动区域和非运动区域在时间方向上进行不同强度的平滑处理;最后对平滑过的深度视频进行重建.实验结果表明,提出的处理方法在保证虚拟视点绘制质量的同时,能够节省14.89%~47.48%的码率.Inconsistency of the depth video deteriorates the quality of rendered virtual view and coding performance.To improve the consistency problem,a novel depth processing algorithm based on temporalhorizontal transposing is proposed in this paper.Firstly,the motioning objects are detected.Then,the depth video and the corresponding color video are transposed towards temporal and horizontal directions.Finally,in the transposed domain,the depth values in motion regions and non-motion regions are smoothened by different intensity.Experimental results show that the proposed algorithm can reduce the bit rate ranging 14.89%-47.48% without compromising the virtual view rendering quality.

关 键 词：深度视频 运动对象 时空变换 深度视频处理 

分 类 号：TN919.81[电子电信—通信与信息系统]