Artistry on a Glowing Screen

By Philip Elmer-DeWitt

A field of grass sways in the wind, each blade clearly defined in yellow and green. A molecule of DNA, its 65,000 atoms represented by gleaming spheres, twists and folds into a thick, knotty ring. Oversize baseballs zoom by at impossible speeds, trailed by surrealistic soda fountains and eerily chattering teeth.

These remarkable images, as sharp as photographs, yet as free from the bounds of reality as an animated cartoon, were crafted not by humans but by computers. They were among the 35 video and film clips showcased last week at the twelfth annual gathering of the Association for Computing Machinery's Special Interest Group on Computer Graphics (SIGGRAPH), the equivalent of the Academy Awards for 25,000 artists, programmers and electronics engineers involved in the complex business of making computer art.

For two nights, the wide screens at San Francisco's Moscone Center glowed with some of the most sophisticated computer animations yet produced: TV commercials that showed Chevy vans floating in clouds and Norelco shavers zipping around a racetrack; television network logos replete with spinning globes and sparkling call letters; scientific simulations displaying molecules at magnifications no microscope could achieve; and animal, vegetable and mineral objects more realistically portrayed than ever before. Says Computer Artist William Reeves of Lucasfilm, who created the image of windblown grass he calls Blowin' in the Wind: "I'm not going to claim it's just like nature, but I'm pushing in that direction."

Indeed, creating objects that have natural-looking textures is one of the key challenges facing computer artists. "It's very easy to make something look smooth, like plastic or ice," says Wayne Carlson, director of production at Cranston/Csuri Productions in Columbus. "What's difficult is to give something the mottled look of bark, leaves or grass." Texture mapping, a computer technique akin to wrapping a photograph of a rough rock around a smooth stone, is one solution to the problem. Another involves the use of a class of equations called fractals. "It's a technology for filling in random surfaces in a way that mimics the way nature is random," explains Lucasfilm Researcher Robert Cook. "You want a rock to look like a rock, a random rock."

Although the results may resemble photographs or surrealistic paintings, the methods used require more mathematics than artistry. A computer artist must consider every aspect of what he wants his finished work to be and translate it into a numerical representation that the machine can comprehend.

The basic techniques by which this translation is accomplished were laid out in the late '60s and early '70s by two University of Utah professors, Ivan Sutherland and David Evans, in fulfillment of a contract for the U.S. Department of Defense. Their task: to build a flight simulator for pilot training that would show on a screen the same unfolding landscape the pilot would see from the air. To do this, the Utah scientists first had to program into the computer a precise mathematical model of every tree, house and mountain in the flight path. Then they instructed the machine to put each of those objects into three-dimensional perspective, to give it the illusion of depth and to eliminate those surfaces that would be hidden from the viewer's line of sight.

Today's most advanced graphics systems take Evans and Sutherland's procedure one step further. Using a programming technique known as ray tracing, they follow the path of each ray of light as it travels from its source, say the sun, to the viewer's eye. Upon striking a surface, each ray will be absorbed, reflected or transmitted in accordance with the laws of optics. Programmed with a mathematical model of the behavior of light rays, the machines can re-create lighting effects of dizzying complexity. Caltech's Jim Kajiya, for example, has used ray tracing to show how ripples travel through a reflecting pool.

Once the color and intensity of each point of light have been calculated, those data are converted directly into the pixels, or picture elements, that make up the images on the computer's screen. Each pixel is either red, green or blue. When viewed from a distance, however, they coalesce like the dots in a pointillist painting. Says Lucasfilm's Cook: "It's like mixing paint. If you stand back, they all blend together."

The whole process consumes vast quantities of computer time. One minute of film may involve as many as 100 billion calculations, driving the costs of TV commercials as high as $4,000 per sec. But conventional filmmaking techniques can be even more expensive. Using a Cray X-MP supercomputer and the latest graphics technology, the special-effects team at Digital Productions was able to create the battling spaceships in the film The Last Starfighter for $4 million. To produce the same scenes with scale-model miniatures would have cost $12 million to $24 million.

With bit parts in such movie hits as Ghostbusters and Amadeus, computer graphics has become part of the standard repertoire of filmmakers and art directors. But the technology is not restricted to Hollywood and Madison Avenue. The Defense Department has invested heavily in complex simulators that use computer graphics for training personnel in the use of tanks, jets and submarines. In the simulators that help teach pilots to fly the CH-53 Sea Stallion helicopter, the computer graphics alone can cost $3 million.

The technology is fast spreading to other professions as well. By mathematically building their scale models within a computer, architects can see what large buildings will look like from the ground, the air or the window of a high-rise across the street. Petroleum engineers can explore graphic versions of geological formations thousands of feet below the ocean floor without drilling. Physicians, manipulating the images produced by CAT scanners, can visually probe the brains of patients without having to perform exploratory surgery. Says Don Greenberg, director of computer graphics at Cornell University: "It's like having a doctor walk on the inside of the skull."

Perhaps the most dramatic use of the technology is in the design of new chemicals and drugs. By showing how the molecules of the active ingredients in a drug attach themselves, like keys in locks, to target molecules in the body, computer models can help researchers see how a change in molecular structure will affect a drug's behavior. Says Robert Langridge, who heads the computer-graphics laboratory at the University of California, San Francisco: "I call it computer-assisted insight."

Last week's SIGGRAPH attendees got a taste of that insight during a 3-min. film sequence, produced at Lawrence Livermore Labs, that showed in a few seconds what biology teachers have labored for years to make clear: the precise mechanism by which molecules of DNA fold upon themselves to form thick strands of chromosomes. "It's something you could never do with a camera," says Livermore's Nelson Max. The audience at SIGGRAPH greeted his technological tour de force with enthusiastic applause.

But the biggest ovations last week were reserved for a more subtle use of computer-graphics technology: a touching, 5-min. animation titled Tony de Peltrie. Created by a design team from the University of Montreal, it depicts a once famous musician who sits at a grand piano in the middle of a hardwood floor, tickling the keys and tapping his white leather shoes to the beat of his memories. In striking contrast to the awkward, robot-like characters in earlier computer films, De Peltrie looks and acts human; his fingers and facial expressions are soft, lifelike and wonderfully appealing. In creating De Peltrie, the Montreal team may have achieved a breakthrough: a digitized character with whom a human audience can identify. --By Philip Elmer-DeWitt. Reported by Thomas McCarroll/New York and Dick Thompson/San Francisco

With reporting by Reported by Thomas McCarroll/New York, Dick Thompson/San Francisco