Generation of Multiple Depth Images from a Single Depth Map Using Multi-baseline Information

In this paper, we propose an algorithm for generation of multiple depth images from one depth map using multi-baseline information. Although depth information is essential in three-dimensional (3-D) applications, the depth estimation from stereo matching is still time consuming and inaccurate. Since the real-time application like free viewpoint TV (FTV) require multiple depth information, stereo matching algorithms can have limitations in terms of speed and accuracy. Thus, it is necessary to construct an effective and fast depth map generation algorithm. This paper assumes that there is an active sensor depth camera, which can obtain a depth map in real-time, and supportive multiple cameras. In order to generate multiple depth maps, we use a single depth map acquired from the depth camera and geometric relationships between cameras. Therefore, the proposed algorithm is faster and more accurate than other methods based on traditional stereo matching. Finally, our method is robust to inaccurate camera parameters since it uses baseline information only.