3DTV video synthesis using depth cameras and color cameras

Novel view synthesis within 3D multiview imaging system has become a key task for offering depth perception, and/or providing the ability to freely change the viewpoint position and viewpoint direction. In that sense multiview video (MVV) system has gained significant interest recently, notably for its capacity to synthesis novel views through Depth Image Based Rendering (DIBR). However, the rendering process does not allow to create perfect novel views in general, it is still prone to errors in particular from the 3D-image-warping- induced disocclusions and the coding/transmission of the depth video, which is a key side information in 3DTV video synthesis. In this paper, we introduce our recent researches related to multiview video analysis for enhancing 3DTV video synthesis quality within a multiple color and depth camera framework. In particular we focus on the depth video pre-processing and a GPU-based post-processing that speeds-up the rendering of multiple novel views, in order to achieve satisfactory perceived image quality and viewing comfort.