Global-background based view synthesis approach for multi-view video

Virtual view synthesis is widely used in multi-view video systems. However, artifacts and disocclusion areas along object boundaries caused by discontinuous values of depth map bring a new challenge to view synthesis. In this paper, a Global-background Based View Synthesis (GBVS) approach is proposed and it benefits the multi-view video to synthesize a virtual view when using Depth Image Based Rendering (DIBR). Firstly, inter-frame information is utilized to generate a global-background image. Then this global-background image is applied to complement the images which have been projected to the virtual view. Finally inpainting technology is employed to repair the still exposed regions. Both objective and subjective results illustrate that the proposed system provides a significant improvement in image quality. Additionally, compared with common view interpolation methods which depend on two or more reference views, the presented representation only requires one original view.