The global context for a lanscape is the surface of a planet. This became evident when I began, years ago, to contemplate making fly-bys of some of my realistic three-dimensional places. Planning a fly-by begs the questions: "Where do we come from? Where do we go to?" On the largest scale, we must situate ourselves on the surface of a planet. Also, modeling the proper geometry of optical paths through the atmosphere requires that the atmosphere both lose density with height and curve around the surface of the planet. (Otherwise, we risk generating unrealistic images and/or having problems with unbounded integration of atmospheric effects.)
These concerns led me to develop Gaea, the Earth-like planet you'll see below. Curiosity, chance, and perceived opportunities led me to model several other planets as well.
My ultimate goal is to populate an entire virtual universe with such planets, which you may interactively investigate as if Cap'n Kirk on your own starship Enterprise. A remarkable aspect of these models is that, since they are based on random processes, they are full of serendipity even for their creator -- I have little better idea exactly what will be found there than you, the casual explorer. Thus I call the act of their creation "playing God in a found universe."
While realism and beauty may in some cases have been achieved already, many year's work will be required to bring their rendering times up to interactive rates.
Here we see a laboriously constructed pair of models of an Earth-like planet -- "Gaea" -- and a Moon-like companion -- "Selene". These procedural texture models are described in some detail in my first book "Textures and Modeling: A Procedural Approach" (Academic Press, 1994).
Here is an early rendering of the planet Gaea. The coloring is ... not so interesting, and the terrain and coastlines are monofractal, as opposed to the multifractal model used in "Gaea & Selene." But it was a start.
One problem with the first version of the planetary atmosphere surrounding Gaea is that it is isotropic about the center of the planet. Thus in quarter-lit renderings the atmosphere is as bright on the night side of the planet as it is on the daylit side -- clearly a wrong model. This image was a test of an early scheme to fix this problem.
This is a later scheme to fix the atmospheric shading problem, this time including a continuous darkening of the atmosphere by time of day, and a nice red coloring near the terminator (the line of transition from day to night) representing sunset/sunrise. Both of these atmospheric shadowing models are technically problematic in picayune ways, but they're simple and look good in a static, distant view.
One of my true (and few, I'd like to hope) true Rube Goldberg constructions led to the model of this semblance to Earth's moon. (This model is also described in "Textures and Modeling: A Procedural Approach".) Interestingly, it contributed to the model of the moon developed by Digital Domain for the movie "Apollo 13" starring Tom Hanks.
Click here for a larger image Click here for a larger image
The folks at Digital Domain had seen the 1989 moon, and asked me to help them model the moon for the "Apollo 13" movie. So Larry Gritz and I set out to improve on that model for them. What we came up with was pretty good, but too slow to render for their purposes. We could have made it better, but it would have gotten slower still. In the end, they used the BRDF shader that Larry coded for them, without which no surface is going to look very moon-like.
(A BRDF is a bidirectional reflection distribution function, which determines how light reflects from a surface. The moon has the bizarre BRDF characteristic of a dusty surface, as first described for computer graphics by Jim Blinn. Basically, it's a good retroreflector -- sending light right back in the direction from whence it came -- and has a strong sidelobe component, close to parallel to the surface and without much sensitivity to the illumination direction.)
Larry is now at Pixar, doing lighting for the upcoming Disney movie "Toy Story".
This is one of my favorite images for its singular simplicity: It consists of one fractal bump-mapped white sphere and one fractal bump-mapped silver sphere, raytraced. I find it remarkable that such simple models can provide a visual approximation to a natural scene; that in turn argues strongly for the validity of fractal geometry as a fundamental language of nature.
Venus, as it appears in the ultraviolet wavelengths, can be modeled with a strikingly simple procedural texture (again, see "Textures and Modeling: A Procedural Approach" for details). This simple model gives a very close visual approximation to the appearance of Earth's sister planet.
The beauty of Jupiter is attributable to two things: Its lovely colors and the wonderful swirls in its clouds. The color we can imitate well. The cloud features are much harder, as they are an example of turbulent flow -- one of the great bugaboos of computer graphics to date. I created this image as an accurate reproduction of a Voyager image, replete with Io and the shadow of Ganymede on Jupiter, to illustrate the visual shortcomings of current CG models of turbulence for my doctoral dissertation.
The delicate rings of Saturn, it turns out, can be modeled by simply modulating transparency of a disk with a rotated sweep of a one-dimensional fractal function.
While working on the planetary atmosphere model, I simply used Perlin's "marble" texture with a spectral color map on the planet itself. (Mostly out of sloth, plain and simple.) Some time later, as the Voyager spacecraft approached Neptune and started sending images back, I was dumbfounded and the uncanny resemblance they bore to this knock-off test image I had created years before.
Some low-orbit renderings from my synthetic planets.