"Ohanian and Dubes" Test Suite

This test suite uses the images from a comparison paper by Ohanian and Dubes (1992). The suite comprises four types of images with four examples of each. The image types are fractal, Markov random fields, leather, and painted surface. The MRF and fractal images are artificially generated while the painted surface and leather images are natural textures. Four 4-way tests (one test discriminating the examples within each image type) and one 16-way test (discriminating all 16 example images).

NOTE:Ohanian and Dubes (1992) use a k-nearest neighbour classifier with leave-one-out testing. Leave-one-out testing is achievable with the framework but it can not be done easily. Some modification to the algorithm functionality is required to implement this.

The following note from Professor Dubes was supplied with the images.

---
Placed here at Dr. Dubes' request on 22 June 1993 by John Lees

   The images are those used in the paper "Performance Evaluation
   for four classes of textural features" by Ohanian and Dubes,
   Pattern Recognition v. 25, 1992. Images are described on p. 824 of
   the paper. A total of 16 images were used: 4 fractal, 4 Gaussian
   Markov Random Field, 4 natural images of pieces of leather and 4
   natural images of painted surfaces.

---

Philippe P. Ohanian and Richard C. Dubes (1992), "Performance Evaluation for Four Classes of Textural Features", Pattern Recognition, v25(8), pp819-833