Example-Based Depth Generation from Single Image for 3D Content
Kai‐Che Liu, Qi Wu, Wenchao Chen, Cheng-Feng Wu, Fu-Chiang Jan, Tsuhan Chen · 2008
2D plus depth is the most convenient way of representing 3D content. Unlike traditional methods of depth map generation which recovers precise geometry from multiple images, this paper presents an example-based method for automatically creating a layered 3D depth map from a single photograph. In our method, example-based background matching is used to identify the closest background and segment the foreground. Then, we do the foreground objects detection and project these objects into perspective using their location. In the end, we provide a thorough quantitative result of our algorithm on a set of real photos which confirms the benefits of our integrated approach. Therefore, a single image can be the content for 3D world.