Conservative visibility preprocessing for efficient walkthroughs of 3D scenes
Year: 1993
Type: article
Abstract: In this paper I address the problem of pre-computing visibility information to increase the efficiency of interactive walkthroughs of large 3-D CAD models such as buildings. The approach presented assumes the use of graphics hardware including hidden-surface removal hardware. The goal is to automatically partition viewpoint space into regions and pre-compute visibility information for each region that is conservative in the sense that the display list computed for a region is a superset of the set of objects that are weakly visible from that region. To render a view from a region, the display list for the region is processed with hidden-surface removal hardware, so that the correct scene is rendered. However, only a small fraction of the faces in a scene may have to be rendered from any viewpoint. The conservative nature of the visibility computation makes possible the use of many techniques for speeding the preprocessing phase and reducing storage requirements for visibility information, including the partitioning of scene faces into objects and the selection of large scene faces (called walls) for occlusion tests. The amount of time and space required for the visibility pre-computation depends only on the number of objects and walls and not the size of the scene, so it is possible to adjust the amount of pre-computation time and space used, with corresponding effects on the efficiency of the on-line walkthrough. Results concerning the number of faces that must be rendered under various object selections are presented for an example model.
DOI: 10.20380/gi1993.25
Author: Plantinga, Harry
Institution: University of Pittsburgh
Language: English
Field: Computer Science
Domain: Physical Sciences
Open Access status: bronze
OpenAlex ID: W322234717
