Click here for a larger image
When I came to work with Benoit at Yale, the first task he gave me was of attempting to generate rivers with the hexagon-subdivision fBm generation scheme described in Appendix A of "The Science of Fractal Images." This image marks my best attempt to make these too-regular models look like a natural landscape. The sedimentary rock texture seen here was later used in "Lethe" and in "Misty".
To my considerable surprise, some of those who've liked this image best were geologists.
Click here for a larger image
I wondered what a black mist in a valley would look like. This was the result.
I wish I could reproduce this image now, antialiased and at higher res. But alas, it belongs in the Journal of Irreproducible Results.
Click here for a larger image
When I invented the grid tracing scheme, I wanted to see if I could ray trace a scene with over a million distinct, non-repeated triangles. This image was the result -- it worked!
To my chagrin, this test-image/knock-off was my mother's favorite among all my images.
Click here for a larger image
We worked hard to get rivers out of our hexagon-subdivision fBm generation scheme. But it just wasn't working out like we'd hoped. In a fit of pique I decided to shove the whole thing under water and see if it it would make a decent coral head.
Exactly one person I ever showed this image to, saw it as what I intended. Most people think it's some kind of lame cloud.
But it's got pixel-sized triangles; millions of 'em...
Frustration with the hexagon-subdivision method for creating rivers (which others have since tackled with very nice results) lead me to team up with Rob Mace in developing his idea for physical, dynamic erosion models. Mandelbrot despised these models -- "Not enough bang for the buck!" -- but they're an area of research I'm still pursuing. (Turns out they involve some particularly nasty PDEs and therefore some real numerical analysis, which I've been work on with Ed Bolton of Yale Geophysics.)
Last Change: March 10 1995
