The computer graphics image "RD Texture Button" by Andrew Witkin and Michael Kass is a milestone in the esthetics of technology. A scientific problem has been rendered with technological means creating such wonderful structures that they were put together into a collage-like image.

'RD Texture Buttons' is a collage of patterns that were created by using computers to simulate a process called reaction-diffusion in which several chemical or biological substances compete for dominance according to simple local rules. The rules govern the formation and decay of each chemical and the way they spread out. This strictly local process can give rise to a surprisingly rich set of global patterns.

In 1952, Alan Turing proposed reaction-diffusion as a mechanism to account for biological pattern formation during embryonic development. Since then, specific reaction-diffusion models have been invoked to account for the formation of zebra stripes, the coloration of sea shells, the markings on butterfly wings, and even the organization of the brain.

Our own exploration of reaction-diffusion patterns began with an interest in the formation of fingerprints. As a test of our model, we began with an image of a real fingerprint, excised a square, and used reaction-diffusion to grow back the missing piece. Although the regrown piece differs from the original in fine detail, in overall appearance it is very much the same.

In addition to imitating patterns that occur in nature, the reaction-diffusion process is capable of generating patterns that are entirely new and unfamiliar yet look strikingly organic, as if they could be drawn from nature. "RD Texture Buttons" explores these possibilities. We produced all the patterns using the same basic mechanism by adjusting the number of simulated chemicals, the local interaction rules, and the manner in which the chemical concentrations are rendered.

One of the most important local interaction rules governs the rate at which chemicals spread out. By varying the rate with direction and location, we produced stripes as well as radial and concentric structures. We created more elaborate woven patterns by increasing the number of interacting chemicals.

To produce the final image, we used our own software running on SGI 4D workstations to simulate the reaction-diffusion system. The immediate result of the simulation was a set of arrays of numeric values representing the final chemical concentrations. We rendered the image using Photorealistic Renderman, defining Renderman shaders to modulate the buttons' color and shape as a function of the chemical concentrations. (Andrew Witkin/Michael Kass)

Technical Background

HW: SG 4D / 210
SW: In-House, Renderman

Andrew Witkin (USA), Professor of Computer Science at Carnegie Mellon University received his B.A. from Columbia College and his Ph.D. from M.I.T. Before joining the faculty at Carnegie Mellon, he headed the perception and graphics group at Schlumberger Palo Alto Research. He works in the areas of computer graphics and computer vision and is the author of several award-winning technical papers. His previous computer graphics works include the animation "Knot Reel" which received the Grand Prix at Parigraph '87.

Michael Kass (USA) is a Staff Research Scientist with the Advanced Technology Group of Apple Computer working on physically-based techniques in computer graphics. He received a B.A. from Princeton University, an M.S. from MIT, and a Ph.D. from Stanford University. Before joining Apple Computer in 1988, he worked at Schlumberger Palo Alto Research in the field of computer graphics and computer vision. He is the author of several award winning technical papers. His previous computer graphics works include the animations "Knot Reel", which received the Grand Prix at Parigraph '87, and "Splash Dance" which received the Grand Prix at Imagina '91.