欢迎访问中国科学院大学学报,今天是

中国科学院大学学报 ›› 2010, Vol. 27 ›› Issue (5): 638-644.DOI: 10.7523/j.issn.2095-6134.2010.5.009

• 论文 • 上一篇    下一篇

基于深度图像的视点绘制新方法

郁理, 郭立, 袁红星   

  1. 中国科学技术大学电子科学与技术系, 合肥 230027
  • 收稿日期:2010-02-05 修回日期:2010-04-12 发布日期:2010-09-15
  • 通讯作者: 郁理
  • 基金资助:

    教育部-英特尔信息技术专项科研基金(MOE-INTEL-08-08)资助 

A novel method of depth-image-based view synthesis

YU Li, GUO Li, YUAN Hong-Xing   

  1. Department of Electronic Science and Technology, University of Science and Technology of China, Hefei 230027, China
  • Received:2010-02-05 Revised:2010-04-12 Published:2010-09-15

摘要:

针对虚拟视点合成中存在的伪影和空洞问题,提出一种基于深度图像的视点绘制新方法.该方法首先使用视点同步生成机制得到虚拟视点的图像和深度信息,并根据深度信息擦除背景伪影;其次,通过基于深度的空洞填补和边界处理方法进一步消除视点图像中的空洞和前景边缘失真.实验结果表明,与MPEG 的3DV/FTV标准参考方法相比,该方法生成的虚拟视点PSNR值提高了2dB.

关键词: 视点合成, 3D变换, 基于深度图像的绘制, 3D视频/自由视点电视

Abstract:

Against the ghosting and holes problem, this study proposes a novel method of depth-image-based view synthesis. First, we present a simultaneous rendering mechanism to generate the depth and image of virtual view at the same time, and then erase the ghosting by the depth comparison. Secondly, by means of depth-based hole filling and boundary treatment, we remove the holes in virtual view and boundary distortion. Experimental results show that, the PSNR of virtual view using this method is 2dB higher than that using MPEG-3DV/FTV reference method.

Key words: view synthesis, 3D warping, depth-image-based rendering, 3D video/free-view TV

中图分类号: