Point visibility queries are a well known problem in computational
geometry. They are also a major building block of photo-realistic
image synthesis. Despite their simple definition: find out whether two
points are mutually visible and if not find the closest obstacle to
the first point, efficient point queries are still an open problem.
For repetitive queries on static geometry, efficiency is achieved
through spatial index structures.
In this talk we present a short overview of the common spatial index
structures in computer graphics, with emphasis on construction
algorithms and the trade-off between construction speed and query
speed. We focus on Bounding Volume Hierarchies, an spatial index
structure regarded currently as most practical, with respect to
computer graphics. In the main part of the talk we discuss several new
and unpublished approaches for query optimized BVHs. Finally, we show
practical results for one of them.