Environment Walkthrough by Overlapped Cull-Boxes

By replacing far geometry with depth-images, Cull-boxes can improve the performance of the walkthrough system. This paper presents the overlapped Cull-box approach. In the model space this paper places overlapped and same-size Cull-boxes, which partition each other into Sub-Cull-boxes. The geometric data is organized by Sub-Cull-boxes, which facilitates geometric data pre-fetching and rendering. Cull-box selection is optimized by evaluating the viewing distance and viewing direction, which ensures more geometric rendering and better image quality. To eliminate holes in depth-image reconstruction, front and back Cull-boxes' depth-images are utilized. A simple and efficient approximate culling algorithm for depth-image rendering is presented. By restricting the viewing angle, the depth-images to be output can be dramatically minimized. The approach has been verified by a prototype system.