Using Reaction Diffusion for Image Processing

In the Summer of 1999 I have experimented with programming COP's (Composite OPerator) for Side Effects' Houdini. These experiments are based on two siggraph papers [Turk 1991, Witkin 1991].

What I did, was use three chemicals as the input for the iterative reaction diffusion loops: Red, Green and Blue. It needed quite a bit of tweeking, since one of the color channels always ended up either fully zero (black) or one (white). Even in the finished COP I wrote for Houdini, the setting of input parameters is critical. One of the colors always seems to dominate, so I made that a choice in the parameters (dominant color).

Original Image Processed Processed: Detail See the separate color channels: RGB Red: Notice the pattern that evolved! Green Blue

And yes, since it is an iterative process that takes at least a hundred iterations to start forming patterns, my Reaction Diffusion Cop worked pretty slow. Nevertheless, I did experiment with using it for animation a bit.

References:

• Turk, Greg: "Generating Textures on Arbitrary Surfaces Using Reaction Diffusion", Computer Graphics, Volume 25, Number 4, July 1991.
  
• Witkin, Andrew, and M. Kass: "Reaction-Diffusion Textures", Computer Graphics, Volume 25, Number 4, July 1991.

^ TOP ^