A view synthesis method based on object coding and simple disparity model

Abstract

We propose a simple view synthesis method based on object coding and simple modeling of disparity space. In the proposed method, a scene from a binocular stereo-vision camera is divided into objects and planes. Our purpose is not a "physically correct" but a "subjectively acceptable" view synthesis within the video-rate, in order to provide convincing images, using MPEG4 object video coding. The proposed method uses interpolation/extrapolation to generate images from arbitrary viewpoints. It also conceals void areas behind objects by using surrounding texture information.