(英) |
International standardization of multi-view and 3D video coding formats is now the hot topic at MPEG. For the latter, video capture by an RGB-D camera is assumed, and video signals and depth maps are utilized simultaneously for efficient compression. Moreover, video sequence with a depth map can be taken by a camera with RGB-D (depth) sensor like Kinect. In this paper, Experiments show that further improvement of picture quality for H.265/HEVC compressed streams. Concretely, we separate video signals into foreground and background, and detect the position of foreground regions by using this information. Furthermore, we detect boundary regions of video signals by using depth map, and we achieve to improve H.265/HEVC. |