Object removal in multi-view images using disparity layering and piecewise homography

Abstract

In this paper, object removal in multi-view images is considered where occluded image information is synthesized from other views through disparity estimation to build layered depth maps and then a piecewise homography is used to transform occluded information from other views. Experimental results show improved performance over single homography transforms in synthesizing the occluded image information after the object is removed.