The Brodatz textures are the most commonly used texture data set, especially in the computer vision and signal processing community. Because they are so commonly used by previous texture analysis/synthesis works, it is almost inevitable to include at least some of them in a texture synthesis paper. In my submitted paper I do use 5 Brodatz textures and 5 Vistex textures for the results section, for the purpose of backward compatibility with earlier papers.

Despite the popularity, however, Brodatz textures are actually copyrighted. I heard some rumors about that they are royalty free, but they are actually NOT. If you go grab a copy of the original book (there is a revised version published in 1999), you can find the following copyright text:

This book belongs to the Dover Pictorial Archive Series. You may use the designs and illustrations for graphics and crafts applications, free and without special permission, provided that you include no more than three in the same pulication or project. (For permission for additional use, please write to Permissions Department, Dover Publications, Inc. 31 East 2nd Street, Mineola, N.Y. 11501)

However, republication or reproduction of any illustrations by any other graphic service, whether it be in a book or in any other design resource, is strictly prohibited.

I am not sure how they deal with research papers now, but to be safe, it is better to include at most 3 texture images in a paper. That is why I used 10 Vistex textures in the result section, and use only 3 Brodatz textures in the TSVQ section and constrained synthesis section. It actually turns out to be a better solution, since the SIGGRAPH proceedings print every page in color anyway.

I have some personal comments about the texture data sets. I believe the Brodatz texture set is indispensible, even with the existence of VisTex. The reason is that most of the Brodatz textures are photographed under controlled lighting conditions, so the images are of very high quality (if you scan them carefully). In addition, they expose the most amount of textures so that irrelevant information such as noise and "non-textural" stuff are not there. Vistex textures are photographed under natural lighting conditions, so they are more tough; i.e. the images contain more noise, and extra visual cues such as shadows, lighting, depth, perspective, etc. So at least seems to me the Brodatz and VisTex textures complment each other.

Another interesting fact to know is that I saw much more usage of Brodatz textures in IEEE publications than ACM ones. This is probably because a lot of major conferences in vision and signal processing are held by IEEE. I don't knwo what will happen if Dover does take some legal actions toward this.

Here are additional information about the copyright issues of Brodatz textures. I got it from Roz Picard.


liyiwei