Visibility as an intrinsic property of geometric models

Abstract

This paper presents a technique to improves the performance of algorithms that exploits visibility based analysis in computer graphics, such as ray tracing. The presented approach maintains a list of visible polygons for each and every one of the polygons in the model as well as for optionally, the light sources in the scene. A pre-processing visibility analysis stage that is view-independent is computed once per scene, creating a Visibility Data Structure (VDS) that becomes part of the model. This, before any visibility intensive computer graphics algorithm takes place. The approach presented may be combined with most acceleration methods for ray tracing such as octree or voxel based spatial subdivision. Furthermore, any algorithm that requires visibility computations can exploit the VDS, from Radiosity to NC machining. Though the presented technique is fairly intuitive, it demonstrates the usability of a visibility data structure of a scene.