A hybrid matching algorithm based on contour and motion information for depth estimation

Abstract

Depth estimation is an important part of free viewpoint television (FTV) because accuracy of depth information affects directly the synthesized video quality at virtual viewpoint. However, generating an accurate depth map is a computationally complex process, which makes real-time implementation challenging. In order to obtain accurate depth information with low complexity, a hybrid matching technique is proposed. It is composed of three types of matching mode: pixel matching, 3x3 and 5x5 block matching. By using contour and motion information in matching mode selection, the proposed technique is more complied with human visual system which is more sensitive moving regions than static one. Experimental results show that the proposed algorithm can not only enhance the synthesized visual quality, but also reduce the complexity in matching process. Copyright © 2013 ACM.