I thought you guys might be interested in helping me gain some insight on this research problem I'm doing at a university: We have an unstructured set of points - sometimes in the hundreds of thousands - acquired from a laser scanner that represent some geometry. We have no knowledge of geometry or orientation. Furthermore, there might be noise - erroneous points that don't belong to the geometry. The goal is to reconstruct the surfaces, essentially, so that we can get some normals and light the points, and thus when the points are displayed it won't simply look like a giant blob of pixels. I've found a few graduate papers, and I can make sense of some of the research. When Delaunay Triangulation and Voronoi Diagrams are tossed in, I sort of stare blankly, but a few of the papers don't require them. I was just wondering if any of you guys had any insights into this problem. Cheers, --Brian

I was considering changing my level editor to work in voxels. I still may do this, and if I do, here's the algorithm I'll use :

http://www.cs.rice.edu/~jwarren/papers/dmc.pdf