A POINT-BASED ASYNCHRONOUS REMOTE VISUALIZATION FRAMEWORK FOR REAL-TIME VIRTUAL REALITY
Jinghua Ge, Daniel J. Sandin, Tom Peterka, Robert L. Kooima, Javier Girado, Andrew E. Johnson · International Journal of Image and Graphics · 2008
High speed interactive virtual reality (VR) exploration of scientific datasets is a challenge when the visualization is computationally expensive. This paper presents a point-based remote visualization pipeline for real-time virtual reality (VR) with asynchronous client-server coupling. Steered by the client-end frustum request, the remote server samples the original dataset into 3D point samples and sends them back to the client for view updating. From every view updating frame, the client incrementally builds up a point-based geometry under an octree-based space partition hierarchy. At every view-reconstruction frame, the client continuously splats the available points onto the screen with efficient occlusion culling and view-dependent level of detail (LOD) control. An experimental visualization framework with a server-end computer cluster and a client-end head-tracked autostereo VR desktop display is used to visualize large-scale mesh datasets and ray-traced 4D Julia set datasets. The overall performance of the VR view reconstruction is about 15 fps and independent of the original dataset complexity.