Materials Library

Greg, how do you judge the difficulties of using the CUReT data to start a materials library? Is this a plausible project, or is it likely to run into serious problems?

Randolph

Ref: <http://www1.cs.columbia.edu/CAVE/software/curet/index.php>

Hi Randolph,

The CURET database is an impressive resource, with about 60 measured materials. This was the original bidirectional texture function measurement work undertaken by Kristin Dana during her Ph.D. studies. I believe she plans a follow-up to this, but I don't know the schedule. Another good resource is the Bonn BTF database, which though it only offers a few surfaces (10 by my count), they are captured at more angles, and the data are pre-aligned, making their application much simpler:

  http://btf.cs.uni-bonn.de/

However, adapting BTF data to a Radiance model is something I've never tried, and my guess is that the memory requirements would kill you. Most researchers try for some kind of PCA (principal component analysis) or tensor representation to reduce the memory footprint. For example, see:

  http://btf.cs.uni-bonn.de/documentation.html
and
  http://mrl.nyu.edu/~alex/research_index.html

How you would incorporate a general Monte Carlo sampling algorithm is a big question. Bottom line, if all you want is a set of materials to use in Radiance, I don't think a BTF representation is the shortest path.

Cheers,
-Greg

···

From: R Fritz <[email protected]>
Date: May 1, 2008 2:17:13 PM PDT

Greg, how do you judge the difficulties of using the CUReT data to start a materials library? Is this a plausible project, or is it likely to run into serious problems?

Randolph

Ref: <http://www1.cs.columbia.edu/CAVE/software/curet/index.php>

Oh, and while we're on the topic, I should mention that I did some work recently with Mashhuda Glencross and others at the University of Manchester developing a technique for gathering combined height field and texture data using an ordinary digital camera:

  http://aig.cs.man.ac.uk/research/daedalus/hallucination.php

I plan to improve upon the software (nasty C-shell scripts) to make the whole process simpler, but it would be interesting to see if this method could be used with the Radiance mesh primitive to tile a convincing surface.

-Greg