(L) [2006/08/03] [Phantom] [Visibility field] Wayback!Stumbled upon this:
[LINK http://www.cs.ucl.ac.uk/research/vr/Projects/VLF/Media/vlfRTpaper/index.htm]
Quite interesting. The author claims that his approach is quite a bit faster than coherent ray tracing, but that result is hard to verify as he only implemented single-ray traversal through a kd-tree constructed with median splits. The presented approach uses a LOT of memory, but it sounds interesting for low polycounts.
_________________
--------------------------------------------------------------
Whatever
(L) [2006/08/04] [lycium] [Visibility field] Wayback!a buddy of mine from my old university (now at nvidia), shaun nirenstein, wrote a paper on doing 5d csg in plücker space to determine exact from-region visibility, which i believe is more general than from-point visibility (ie a single test can be used to cull out visibility calculations in area light calculations).
it also took a ton of memory and ages to precompute, but some other students worked with him to parallelise it and attempt some realtime ray tracing. i don't think they were well aware of the state of the art at the time though (havran's sah), even though shaun was, so direct comparisons between the two were unfortunately not made.
it should be easy to google in case anyone is interested in this variant of ray classification.