Does anyone have any experience with photographically scanning materials to get a light-neutral diffuse/albedo, normal map, or a more robust full BRDF scan?
I've seen an old Ryan Clark tutorial here:
http://zarria.net/nrmphoto/nrmphoto.html
And I'm curious as to whether anyone has set up such a system for a more production-ready setup. The notion of being able to scan material samples from a hardware store (tiles, pavers and bricks, plywood, etc) seems like it would be pretty useful as a production tool for generating accurate, high fidelity materials.
Replies
There was a tool posted on here awhile back that you would feed two photos of a subject, and it would spit out the mesh and a combined photo. Ah, here it is... PhotoSculpt
http://www.polycount.com/forum/showthread.php?t=70506
There are other similar solutions out there, search for photogrammetry if interested.
You would have to remove shading from the diffuse, Crazy Bump kind of does this but it's nowhere near perfect. If you shot the surface inside a light tent, that might help.
But there's the issue of tiling. You would probably want to fix tiling seams on all the maps together at the same time. No easy feat.
I know there are a few contract services that provide this kind of concept, Surface Mimic being one of them, and I was looking for people with experience setting up these sorts of capture rigs. There are a number of research papers about various capture methods but I'm currently having a bit of a problem sorting out the practical from the theoretical and isolating easily-duplicated material scans from the number of 3d partial and/or low quality 3d scan methods.