The restoration game

Michelangelo's masterpiece the Florentine Pieta is unique. Unusually, he began the work without commission, making it a rare piece of self-expression, one of the first in the history of art. This colossal piece of carved marble, at 2.5 metres tall and weighing eight tons, was intended to crown the altar under which Michelangelo wanted to be buried. Then, a few years before his death, he took a hammer to the work to destroy it. A student rescued the piece, but left art historians with a mystery of composition, reconstruction and workmanship. Only today, with the assistance of state- of-the-art, so-called "deep computing", is the eminent Renaissance scholar Professor Jack Wasserman beginning to unravel its secrets. His work will enhance our understanding of the beliefs and genius of this great artist.

The Florentine Pieta is a group of four larger-than-life figures: the broken body of Christ from the cross is held up by Mary Magdalen aided by Nicodemus above her and the Virgin Mary to the right. Only the figures of Mary Magdalen and Christ are finished, although the left arm of Jesus has been broken and repaired and the left leg is missing. After Michelangelo's act of violence against it the otherwise undistinguished sculptor Tiberio Calcagni reassembled it. This, as Wasserman explains, makes it a statue by two artists and even more difficult to interpret as a result.

Michelangelo's style has occupied the minds of art historians since the 16th century. The proportion and details of the Florentine Pieta are particularly curious, since some parts seem elongated while others are suspiciously small. More fundamentally, it has not been possible to view the statue as Michelangelo intended, with the base at eye level: since being rescued it sits on the floor in the Museum of the Opera del Duomo in Florence cathedral.

To tackle these issues, Wasserman originally engaged a photographer to conduct a complete study of the work, as well as an Italian research institute to carry out a scientific analysis. In 1997, his search brought him to IBM's Gabriel Taubin, head of the Visual Technologies Group at IBM Research, and a specialist in geometric computation. Shortly after, the IBM Foundation agreed to fund a project to generate an ultra-high-resolution virtual image of the work.

Not only was the task of artistic interest, but it presented great technological challenges. Digitising the Florentine Pieta for a near perfect replica required taking some 10,000 digital images. A camera originally designed for plastic surgeons to take three-dimensional photographs was modified for the project. It beams stripes of light on to the surface of the object and then takes 12 images simultaneously; six in colour to capture details of texture and shade, and six in black and white. A grid of laser beams is also projected on to the statue to generate more spatial information. From the raw data, a computer algorithm then virtually reconstructs the piece.

Dr Fausto Bernardini, head of the IBM Pieta team, explains that advances in technology were required in several areas. The fundamental challenge was to manage a virtual study of an object of this size. After a test- run on a papier mache mock-up of the statue in the Thomas J Watson Research Center in Yorktown Heights, New York, the IBM researchers went to Florence early last year, where they spent 12 days working in the small, unheated room in which the statue is housed. The end result is a virtual statue constructed of a mosaic of some 10 million triangles, with an additional 9 billion bytes of data refining the geometry and adding detailed colour and texture information.

"Handling this amount of information on a PC was the first advance," Dr Bernardini says. "It would be straightforward on a super-computer, but fast-mesh computation at speed and with compression is a hard task."

The PC viewer first displays a simplified model of the full statue which can be viewed from any angle. Users can then zoom in on any part of the statue in detail down to a single millimetre. "Large portions of the statue are unfinished," says Professor Wasserman. "You can see the sculptors' tool marks, which differ on different parts of the piece. The detail available makes a very rich area of research for the historian.

"The ability to view separately the limbs of each figure that Michelangelo removed would provide valuable insight into the question of how Michelangelo viewed the statue after he destroyed it, and perhaps what his motivation was," he adds.

The final version of the Pieta model is due to be completed this spring, although Professor Wasserman has already been able to draw some conclusions on Michelangelo's work. He has confirmed the hypothesis that he intended the statue for his tomb monument as the adjusted proportions of the figures would appear more realistic once the Pieta was mounted on an altar above the grave. Professor Wasserman also believes that from this angle the piece gives a naturalistic depiction of the Deposition of Christ rather than a symbolic one intended primarily to convey points of theology. This suggestion supports an interpretation of the work as self-expressive of Michelangelo, revealing of the artist's piety and even psychological state at the end of his life.

Being able to handle large amounts of data in near real-time to draw out results immediately available to end-users has led to the coining of the term "deep computing" by IBM. Dr Bernardini believes that their work will find application well beyond the art history arena. "The creation of 3-D content is expensive and requires experts. But these advances mean a 3-D camcorder might evolve that builds up an image automatically," he explains.

In 1553, Asconio Condivi, author of The Life of Michelangelo, wrote of the Florentine Pieta: "Let it suffice: I tell you it is a rare thing, and one of the most laborious works that he has yet done." Little would he or Michelangelo have imagined that the masterpiece would be at the centre of advanced computer research, over 400 years later.