Click here for a larger image
In the summer of 1991 I worked at NASA Ames in California at the Visualization for Planetary Exploration Lab. They had a VR system there that allowed one to view, interactively with a head-mounted display, the Martian panoramas sent back by the two Viking landers. They were interested in utilizing my capabilities for terrain imaging in their virtual environments, which are designed to help planetary geologists efficiently extract information from the visual data sent back by planetary probes (in this case, the Viking probes). One means of doing so is to create height fields from stereo pairs taken from orbit, and to put the resulting terrain models into a virtual environment for exploration.
My imaging technology is hardly real-time, so I elected to create a synthetic variation on the Viking panoramas. (The panoramic virtual screen I created for this purpose is documented in "Graphics Gems III," pages 288-294.)
This is an image of the Valles Marineris, a stunning feature on Mars. It is approximately the width of Australia or the continental United States, from end to end. In fact, it could never be seen like this in real life, as it would curve over the horizon of Mars, which is a fairly small planet. It consists of two very straight, parallel valleys, which are distorted into bows in the extreme (370 degree by 180 degree) wide-angle field of view of the panoramic projection. We are situated above the middle of the narrow peninsula separating the two valleys, in this image.
All coloring, surface texture, and atmospherics are strictly artistic, and not driven by empirical data or scientific models. Nevertheless, artful use of a procedural texture on the terrain surface brings out certain subtleties of morphology that illumination alone would not make visible. Thus it may be seen as "art in the service of science."
I also rendered this panoramic view in stereo, which was a little bizarre... (See the Graphics Gems article.)
Click here for a larger image
Mandelbrot's physics group at Yale were studying the growth dynamics of diffusion-limited aggregations (or DLA for short) during my time there. One thing they did was generate 3-D clusters in dynamic simulations of the growth process. In such a simulation, one fixes a single sphere in space, then simulates others moving in a random walk. When any of these wandering spheres touches the cluster, it sticks. After adding 10,000 spheres in such a simulation, one gets a cluster such as this. (Such objects are ubiquitous in Nature, in soot, interstellar dust grains, percolation clusters, fish feces floating to the bottom of the ocean, etc.)
This complex shape is difficult to visualize clearly -- if you simply project it onto the image plane, you get a visual mess. The issue then is to clearly illustrate the spatial structure of the cluster, and to illuminate its growth history with color. (Red spheres were added first; violet spheres last.) I used the atmosphere model I developed for my planets, to disambiguate near from far in this image. Interestingly, there is no light source in this image: The are simply 10,000 shiny spheres and a non-physical model of an atmosphere, which has a property of changing the color of a ray over distance, without emitting any light that would illuminate surfaces. Thus the atmosphere is visible both directly and in reflection from the spheres, giving the illusion of illumination. But remove the atmosphere and this image would be perfectly black! Such can be the fun of synthetic imagery and non-physical modeling.
This same cluster is seen in the artistic rendering "Fractal Mandala".
