Bilayer representation for three dimensional visual communication

This paper proposes a bilayer representation of scene consisting of foreground and background. It uses two layers to represent foreground and background, respectively. Unlike previous layer-based approaches, both layers cover the whole region instead of interested regions. Based on the bilayer representation, bilayer stereo matching, view synthesis approach and foreground/background separation algorithm are developed for three dimensional visual communication. They all consist of two steps: recovery of disparity fields of both layers and composition of these disparity fields; synthesis of images of both layers and composition of these images; computation of two layer costs and separation of foreground and background. Experimental results demonstrate that both high quality and high efficiency are achieved by above approaches. They meet the requirements of desktop-based three dimensional visual communication.