San Jose Mercury News
October 10, 2000
by Lisa M. Krieger
Computer scientists have detected something that centuries of adoring art buffs have not: Michelangelo's sculpture David, an icon of physical beauty, can't see straight.
A laser scan of his face, performed by Stanford Professor Marc Levoy, reveals that his handsome eyes diverge. The right eye stares straight ahead; the left eye looks to the left.
Tourists at the Galleria dell' Accademia in Florence, Italy, will just have to take this on faith. From the ground, a close-up, full-frontal view of David's face is impossible because of his upraised left hand. In profile, when seen from below, each eye looks normal.
''It's almost certainly not a flaw'' but a deliberate artistic device, said Levoy, who discovered the divergence while perched on a scaffold, just yards from the face.
''It is widely accepted that Michelangelo introduced distortions into his statues in order to achieve certain visual effects,'' Levoy said.
Studying the renderings of the 3-D computer model, he created an explanatory hypothesis: ''Michelangelo may have designed the direction of the gaze so that there would be an eye suitable for each profile,'' Levoy said. ''From one profile, he was looking at Goliath, about to do battle. The other side is a noble profile of the future king of Israel, boldly looking straight ahead,'' Levoy said.
The computer model also dramatically displays things not easily detected by mere mortal eyes, such as the thin marble ridge carved to exaggerate the boundary between David's lips and his upper face. Its measurements show that the famous furrowed brows protrude from the forehead in a way that is anatomically impossible. The team's work also revealed that he stands 17 feet tall, not the mere 14 feet 3 inches recorded in art history books.
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, reveals these things -- and much more.
The technology will make it possible to view a sculpture from different angles, zoom in on details as small as a chisel mark, change lighting conditions to see how they affect a statue's appearance, and maybe even animate the figures.
A two-dimensional photo fails to do justice to these priceless works. And replicas of the sculptures are not accurate because they are based on contemporary artists' interpretations.
This ambitious research project aims to bring highly realistic three-dimensional images to display screens at art museums or personal computers. The project -- funded by Stanford University, Interval Research Corp. and the Paul G. Allen Foundation for the Arts -- will eventually create a digital archive of these cultural artifacts.
While laser scanning has been around for a while, techniques for combining multiple images shot from different angles are relatively new. The techniques for this project were invented by Levoy's team several years ago. Moreover, no one has ever scanned large statues at a very fine resolution. The combination of size and resolution is creating data sets several orders of magnitude larger than other 3-D computer models.
''Our goal is to push the state-of-the-art of 3-D scanning, creating models that were 1,000 times larger than ever made before,'' Levoy said. ''Secondly, we want to provide a tool for art historians. Thirdly, we want to provide a permanent archive of these statues in the event, heaven forbid, that something happened to the originals.''
Trained as an architect, Levoy brings to his work the aesthetic training of an artist.
''I remember a semester at Cornell,'' where he attended undergraduate school, ''that I blew off all my classes and wrote a long paper on the art of Michelangelo. Little did I know this would become a lifelong passion in architectural history.''
On a yearlong sabbatical from Stanford, Levoy and his wife traded his Palo Alto condominium for a Florentine villa and placed their children in the Italian public schools.
He shipped over a four-ton scanner in the cargo hold of a Boeing 747 jumbo jet. He brought with him a hard-working research team -- of which 14 of its 32 members were excited Stanford undergrads, each with a year of conversational Italian to their credit.
A palazzo on the Arno River was converted into a temporary computer graphics lab. Scanners were assembled and tested; software was written. For five months last year, the group moved from museum to museum, digitizing artworks such as David and sending the data back to the lab for post-processing. Each time they moved to a different museum, the 1,800-pound gantry had to be re-erected.
To scan David, they worked in two shifts, from 7 p.m. until 8 a.m. when the museum was closed to visitors. A total of 7,000 color images were created.
The divergence of David's eyes and his exaggerated brow line and lip ridges are not the only examples of Michelangelo's artistic distortions, Levoy noted. David's head and right hand are too large for his body.
If the Virgin Mary in The Pieta were to stand up in The Pieta, it would be evident that her legs are extraordinarily long; seated, they gracefully cradle her dying son on her lap. Her right arm, supporting Christ's shoulders, is also too long.
In addition to David, the team scanned Michelangelo's four Unfinished Slaves and St. Matthew. At the nearby Medici Chapels, they scanned his four allegorical statues -- Night, Day, Dawn and Dusk -- as well as the architectural settings of both museums.
Their most recent project has been the digitization of the fragments of the Forma Urbis Romae, a giant map of ancient Rome carved onto marble slabs 18 centuries ago -- and now in fragments.
One of the key source documents of ancient Roman topography, its surface incisions show every street, building, room and staircase in the ancient city. But with the disintegration of the Roman Empire in the fifth century A.D., the marble map was used to build walls or burned in kilns to make lime for cement. Only 15 percent of the original map remains, but this fraction covers a few important areas of the city, with portions of the imperial forums, the Colosseum and the Palatine Hills.
Based on 3-D shape, marble veining and other clues, the pieces of this historic jigsaw project may someday fit together, said Levoy. Scans are now being assembled to create 3-D models of each map fragment.
Now back in Palo Alto, the team is working on building 3-D models from their data. It is a complicated task, made more difficult by the large size of their data sets. For each statue or map fragment, they must clean them up, align them with each other, merge them to form a seamless ''mesh,'' fill holes in this mesh and finally map color photographic data onto it. Thus far,they've built a full-resolution model of the statue of St. Matthew and several medium-resolution models of the head of David.
Computer models will never replace a museum, Levoy said. No model displayed on a computer screen can evoke the emotions of a larger-than-life marble statue.
On the other hand, computers may change the way that people view art, he said. Museum visitors see most statues from a limited set of viewpoints, usually from the ground looking up; computer models give people a way to look at statues from any viewpoint.
They can also change the lighting. Chisel marks, for instance, are more dramatic on a computer screen than on the statue because the viewpoint and the light are different.
The research project has only deepened his respect for Michelangelo's artistry, he said.
He recalled his last night of scanning, perched alone on the scaffold, next to David.
''It was 4 a.m. and I was on eye level. I knew it was my last night,'' Levoy said. ''I just sat there for three hours. As the sun rose, I watched the light change across his face. It was a moving experience.''
IF YOU'RE INTERESTED More information about the Digital Michelangelo Project is at http://graphics.stanford.edu/projects/mich/.
The task of building 3-D models from raw data is only partially complete, but the team has created a skeletal archive of 3-D models and placed its catalog online. The models in the archive are available to qualified scientists. To find out more about obtaining a license, look at the Web page. Commercial use of the models is prohibited. Any scholar who would like to use these models should send an e-mail request to dmich-request@graphics.stanford.edu stating name, title, affiliation, intended use of the model and a statement accepting the terms of licensing.