** Drexel Vision Group: Texture Database **
Over 40,000 images of 20 different textures
-------------------------------------------


About the Database
==================
The Drexel Texture Database contains many images of 20 different textures. Each texture is imaged under a moving light-source at multiple distances, and with different in-plane and out-of-plane rotations.

This database was originally released in conjunction with the following paper. Please cite this work when using our database:

    G. Oxholm, P. Bariya, and K. Nishino,
    "The Scale of Geometric Texture,"
    in Proc. of Twelfth European Conference on Computer Vision ECCV'12, Part I, pp58-71, Oct., 2012 [PDF]

Note that this database is provided for your benefit without warranty or guarantees of any sort. If you note any problem with the database, please email us, but please note that we may not have the resources to deal with all issues.


Technical Detail
================
These photographs were taken in a darkroom under a point light source. A reflective sphere was used to determine the lighting direction, and a white piece of paper was used to account for any variation in the apparent illumination intensity. Raw image data from a single image has been converted into EXR format (multiple exposures were not used to generate each image).

The database contains 20 directories, one for each texture. Each of these directories contains many different viewing condition sub-directories. Each of these sub-directories is given a name as follows:
DX_INYY_OUTZZ
* X is the distance (1, 2, or 3)
* YY is the in-plane rotation in approximate degrees
* ZZ is the out-of-plane rotation in approximate degrees

In each of these directories there are the following files:
* light.txt - A list of lighting directions
  Each row has the form: IMAGE_PREFIX L_X L_Y L_Z L_R L_G L_B>
     - IMAGE_PREFIX - the image filename excluding the extension
     - L_X L_Y L_Z - the X, Y and Z components of the lighting direction
     - L_R L_G L_B - the RGB intensity values for the light. More detail is below.
  Lines starting with a semicolon should be ignored (the image has likely been removed as well). These images were faulty, or corrupted.
* lights.png - A visual description of the lighting directions, for debugging
* _____.exr  - A series of OpenEXR formatted images. Each of these images will have an associated line in light.txt giving its illumination direction


Light Intensity
---------------
Our images were taken under a point light source whose distance may have varied somewhat as it was waved across the image. The result is that the light source has a non-uniform intensity across each set. To measure the intensity, we included a white piece of paper in the imaging process. In our work (cited above), we normalized the input images according to the estimated light source intensity by dividing the image channel by the intensity value for that channel.


Known Issues
============

Since there are over 40,000 images, not all images have been inspected manually. You should inspect each set carefully before using it so that you are aware of any issues. Please, let us know if you find any problems. Also note that some of the image sets have some out-of-focus regions. This usually happens with the out-of-plane rotations, as the texture begins to span a wide rage of depth.