Techniques for Accelerated View-Dependent Mesh Refinement
James Strauss, Amitava Datta · UWA Profiles and Research Repository (University of Western Australia) · 2003
Abstract. View-dependent mesh refinement techniques typically pre-compute a hierarchical data structure that is queried at run-time to produce an approximation of a given mesh. This approximation, or level-of-detail (LOD), can be used in place of the original object so long as the viewpoint does not change. This should result in no loss of visible detail, yet should be less computationally expensive. The approach of existing techniques is to collapse or expand nodes in the hierarchical data structure level by level. We propose a new method for fast generation of view-dependent level-of-details when the frame-to-frame coherence is low, such as when an object moves or rotates at a rapid rate with respect to the viewpoint. Our method is based on two new techniques for aggressive detection of visible parts of the merge tree data structure. The jump split and jump collapse move the active nodes front up or down the tree many generations at a time, reducing the number and expense of iterations required to refine the mesh. 1