Digital Michelangelo Project: Creating Virtual Sculpture

The Digital Michelangelo Project, an ambitious effort to create the first authoritative, 3-D computer archive of the 15th-century Italian artist's most famous sculptures, is laying the groundwork for these capabilities. The project not only holds the promise of making virtual copies of Michelangelo's statues available for appreciation and study around the world, but it also may set a new standard for the computer representation of three-dimensional physical objects.

The Stanford research team conducting the project is working closely with officials from a number of Italian museums and institutions in Florence, Rome and the Vatican who have given them access to these priceless art treasures. In mid-January, the researchers intend to begin scanning Michelangelo's sculptures in the Gallerie dell'Accademia in Florence, including "The Unfinished Slaves" and "The David." In February they are scheduled to move to the Medici Chapel.

Marc Levoy, the associate professor of computer science and electrical engineering at Stanford who conceived of the project, and a research team is spending the 1998-99 academic year in Italy. The team - which consists of research associate Kari Pulli, five graduate and several undergraduate students - is applying state-of-the-art technology to digitize the famous statues. Once they have scanned a sculpture, the researchers will apply advanced algorithms developed at Stanford to convert the digital data into an accurate, three-dimensional computer model. They will then create a matching overlay that contains additional data about surface color and characteristics required to reproduce its image accurately.

Levoy and his colleagues at Stanford's Computer Graphics Laboratory have been developing these algorithms for several years. In 1996, they used the basic technique to create the world's first "3-D fax." They scanned a small statue, created a 3-D computer model of it, and transmitted the model electronically to a stereolithography plant in Southern California that used the information to produce a detailed plastic replica of the original statue.

The Digital Michelangelo Project will produce some of the largest 3-D computer models ever made. It will also combine color and shape information at a level of detail never before attempted.

3-D computer models have a variety of uses

Once computer models of Michelangelo's statues have been created, they will have a variety of uses:

* The virtual sculptures can be viewed on a computer screen from any perspective. Postcards sold at the Accademia Gallery in Florence show a close-up of the head of "The David," but no tourist can see this view because it's 20 feet in the air. If the Gallery includes a virtual model of the sculpture in its planned multimedia center, visitors could admire the sculpting of David's furrowed brow and piercing eyes as closely as they like.

* Computer graphics can be used to "virtually" restore damaged areas of an artwork. It should be possible, for example, to reconstruct what the beard of Michelangelo's sculpture "Moses" in San Pietro in Vincoli, Rome, looked like before it was worn down by the reverent touch of generations of Jewish visitors.

* Through animation, art teachers can illustrate some of the tricks of perspective that Michelangelo used. In his "Pieta," for example, he was the first to solve the problem of making a grown man fit neatly in a woman's lap. But he did so in part by giving Mary extremely long legs. Using animation to have Mary stand up would show this distortion plainly.

* Many ancient Greek statues were originally painted in bright colors. Computer models can be used to illuminate the sculptures with colored light so precisely that it looks as if the sculptures had been painted, even on close examination. With such a light system, visitors could turn on the lights to see how such a statue originally looked and turn them off to appreciate its current appearance. Such a lighting scheme also could be used to project outlines and other guides on damaged statues to aid in the restoration process.

* A detailed computer model can be used to make accurate physical replicas of a sculpture at any scale. Currently, small replicas of important sculpture are not accurate because they generally are based on a contemporary artist's interpretation. Accurate replicas could be used for both research and commercial purposes. But Levoy emphasizes that the goals of his project are entirely scholarly.

The $1.5 million research project is funded by the Interval Research Corporation and the Allen Foundation for the Arts. Founded in 1992 by Microsoft co-founder Paul Allen and Silicon Valley veteran David Liddle, Interval Research is a high technology laboratory that performs research and advanced development. The Allen Foundation is one of Paul Allen's charitable organizations.

"The Digital Michelangelo Project is an ambitious task that dovetails with Interval's research studies in new market opportunities at the intersection of technology and popular culture," Liddle says. "We believe that real-time interactive graphics has immense potential as a more effective tool for communications, education, and especially entertainment."

Computer graphics laboratory set up in Florence

During the fall quarter, the Stanford researchers set up a temporary computer graphics laboratory adjacent to the Stanford Overseas Studies Center in Florence. There the team has been designing, building and testing the three computer-controlled laser scanners and custom-built mechanical frameworks they will use in the scanning process.

The largest scanner is mounted on a 20-foot truss, can measure distances to within a quarter millimeter, and was built to the researchers' specifications by Cyberware Inc. of Monterey, Calif. The second is a bread box-sized scanner mounted at the end of a precision robot arm that is designed to take measurements in hard-to-reach places. The third scanner is a test model that measures distances by the time it takes laser light to travel to an object and bounce back. It was built by Cyra Technologies of Oakland, Calif., and can measure the distance of objects up to 300 feet away with an accuracy of 5 millimeters.

"The David" is three feet taller than art textbooks say

In their preparatory work, the researchers already have discovered a major error in the art history books. These volumes, and even the guidebooks sold at the Galleria dell'Accademia, uniformly agree that "The David" is 434 centimeters (14' 3") tall without its pedestal. But the computer scientists have determined that the famous statue, which sits on top of a six-foot pedestal, is actually 517 centimeters (17') tall.

"I expected our project to be historically significant," Levoy says, "I never expected it to start by correcting the history books."

The statue's extra three feet of altitude created a major problem for the Stanford team. They used the official figures of its height to size their large scanner, so it was too short to reach the top of the actual sculpture. That forced researchers in Florence and at Cyberware's headquarters in Monterey to work through the holidays to extend the gantry by another three feet without compromising accuracy or safety, Levoy says. They were working against a tight deadline because Accademia officials made elaborate plans to hire extra guards and reroute tourist traffic during the scheduled scanning period.

The researchers intend to scan the sculpture "The Unfinished Slaves" first. They will do so with a quarter-millimeter precision. That will allow them to create a computer model that contains about 100 million triangles and provides sufficient detail to show individual chisel marks. The ability to capture chisel marks is particularly important in the case of this set of figures. Part of the sculptures' fascination lies in the fact that they are unfinished and so provide important insights into how the master sculptor worked.

Computer science issues

"From a computer science perspective, there are several issues that we must deal with," Levoy says.

First, he expects the computer model of "The David" to run to tens of gigabytes, and the total for all of models the team intends to create will run into terabytes. They have never dealt with 3-D models of this size before.

Scanning simultaneously for color and shape is also a question mark. All three of their scanners will be equipped with a color camera for this purpose, but the researchers have never attempted to record both range and color at such a fine scale.

Finally, extracting the intrinsic surface reflectance of the statues will be a challenge. Getting the proper value for the surface reflectance is the key to making the computer models look right under different lighting conditions. To do so, the researchers must process the images to remove the effects of shadows and highlights. They have done this in controlled lighting conditions in the laboratory. But doing it "in the field," where the lighting is more variable, may prove more difficult.

"I think we know how to do all of these things, but putting them together will be a challenge," Levoy says.

After completing their work with the sculpture in Florence, the group intends to move their operations to Rome, where Michelangelo's "Pieta" and "Moses," along with "Laocoon," one of the best-known works of Roman art, are on the agenda.

They also will scan the pieces of the "Forma Urbis Romae," an ancient mosaic map of Rome. Levoy intends to use computer techniques to fit together pieces of the broken map. In addition, the group will take on other projects if time permits, such as scanning ancient musical instruments in the Medici Collection, including a Stradivarius violin.

The project has required extensive cooperation from officials at the Italian museums and institutions that house the irreplaceable sculpture, including the Accademia Gallery, Medici Chapel, Superintendency of Fine Arts in Florence, Superintendency of Fine Arts in Rome, and the Vatican Museum.

###

Related material: Michelangelo Project home page -- http://graphics.stanford.edu/projects/mich/