(L) [2011/12/05] [ost
by ypoissant] [Looking for Nearest Neighbor BVH splitting algo] Wayback!I recall an article that was discussing a BVH splitting algorithm that whent along the following lines:
1 - Pick any two primitive
2 - Compute two centroids from those two triangles
3 - Gather two lists of all the primitives that are nearest those two centroids
4 - If the list members changed since the last iteratiop then
4b - Compute two centroids from the members of the two lists
4c - Goto step 3
5 - Otherwise the BVH is splitted at this level. Continue for the next levels.
Does anybody know an article that discusses such BVH splitting method?
(L) [2011/12/06] [ost
by TheSFReader] [Looking for Nearest Neighbor BVH splitting algo] Wayback!(Shadow007 here, just changing my nickname ! Pleased to find a now functional again ompf forum!)
It seems it's not the paper you want, but your algorithm explanation reminded me of the "Fast Agglomerative Clustering" one by Walter; Bala et al :
[LINK http://www.cs.cornell.edu/~kb/publications/IRT08.pdf]
(L) [2011/12/31] [ost
by hobold] [Looking for Nearest Neighbor BVH splitting algo] Wayback!This method seems to be an application of a more general principle called "vector quantization", which I encountered in the context of data compression. These specific buzzwords might help narrow down your search. (And in case you already knew this trivia, then please don't feel offended that I pointed out the obvious.)