3D视频中基于前景的过渡像素深度划归方法

Foreground-based Decision of Transition Pixels Depth for 3D Video

DIBR(Depth Image Based Rendering)算法的提出,使得3D视频只需要一个原始视点的纹理信息及其对应的深度信息就能绘制出新的虚拟视点,因此深度信息的精准与否将直接影响到绘制的虚拟视点质量。由于深度失真经常发生在前景与背景的过渡区域,本文对该区域像素定义为过渡像素,针对这种过渡像素的深度失真问题进行了分析,并提出了一种基于前景的过渡像素深度划归方法。主要思想是:通过阈值判断找出过渡像素的位置,然后将前景的深度值赋予过渡像素。实验表明,这种方法能够有效的减少绘制过程的失真,提高虚拟视点的质量。

The propose of algorithm of DIBR（ Depth Image Based Rendering） makes the novel virtual view be rendered by providing only one original view＇s texture and its depth information. Hence,the quality of virtual view is related to the accurate of depth information. Generally,the depth distortion often occurs in the transition regions between foreground and background. In this paper,pixels in such regions were defined as transition pixels,and an analysis on this issue was also given. Then,the foreground-based method was proposed to decide the depth value of transition pixels for 3D video algorithm. The main idea is that a threshold is firstly used to locate transition pixels and then the depth value of foreground is assigned to them. Experimental results show that such method could reduce the distortion of rendering and improve virtual view＇s quality obviously.