Decoding the past

Computer experts help archaeologists reassemble fragments of an ancient Greek fresco, writes Tom Avril

A MIGHTY volcanic eruption smothered the Mediterranean island of Thera in the 17th century BC, preserving a storeyed trove of cultural and artistic relics from the late Bronze Age.

When it was excavated in modern times, however, much of the trove was like a giant jigsaw puzzle. Greek archaeologists have laboured for decades to reconstruct striking frescoes from pieces that still boast their original creamy whites and deep reds.

Now, scientists from Princeton University have introduced them to tools that are expected to speed up the process considerably: a laser scanner to capture three-dimensional images of the precious fresco fragments, and a computer programme that figures how they fit together.

Digital imaging has been used in a few other instances to decode the ancient past. What’s different is that the Princeton system was designed with off-the-shelf technology for use by non-computer experts — the archaeologists who had been doing the reassembly entirely by hand.

“I think this will greatly help our work,” says Andreas Vlachopoulos, one of the Greek researchers. “We do not have the painter to ask what he or she has painted 3,500 years ago.”

In collaboration with Vlachopoulos and colleague Christos Doumas, the Princeton team reported successful initial results last month at an international computer graphics conference in Los Angeles.

The software sifted through hundreds of random fragments from the island, commonly called Santorini. It correctly identified 10 out of 12 known matches and two additional “new” ones that had eluded the Greek researchers.

The computer also made short work of a fake fresco that the Greek scholars made — and broke — in order to test the system. The software correctly found more than two-thirds of the 253 possible matches.

That’s not nearly perfect, but then neither are the archaeologists who perform the same task by eyeballing fragments spread out on a table.

Tim Weyrich, a project leader who recently left Princeton to teach at University College London, says the goal is merely to supplement the human expertise.

“This is such a delicate, complex task,” Weyrich says. “We are dealing with the world’s experts on this. We are not so overconfident that we think we can write a computer programme that replaces them.”

The effort began with a chance encounter in 2006. Princeton’s dean of faculty, David Dobkin, was in Greece for a board meeting of the university’s Stanley J. Seeger Hellenic Fund, which supports Greek studies. While there, Dobkin got to visit Akrotiri, one of the archaeological sites on Santorini, and he met Doumas and Vlachopoulos.

Dobkin is also a computer science professor, so he asked the Greek scholars whether computers might be useful in their work.

A collaboration was born. Last year Vlachopoulos came to lecture at Princeton, and the computer scientists made several trips to Santorini, exchanging information and developing their high-tech method.

Dobkin and his fellow Princetonians had no particular expertise in archaeology, but the Pittsburgh native has been driven since childhood to collect and organise things — among them snow globes and US pennies, which are piled high in his university office.

The challenge of piecing together ancient Greek fragments excited him.

“This seemed like an interesting real-world problem that would stretch our techniques,” Dobkin says.

So far, it is unclear just how much the software method will accelerate things. In the paper presented at the conference, the authors estimate that the time to reconstruct a complete painted wall might be reduced from years to months.

But certain tasks cannot be rushed. As the fragile fragments are unearthed, for example, they must first be cleaned and stabilised.

And for fragments where the painted decoration can be clearly seen, the computer is sometimes no faster than the human eye, Weyrich says. The advantage comes with the countless pieces that have no distinguishing decorations, just a background of uniform colour.

The laser scanner maps the contours of each piece, sampling thousands of data points. The software then compares the points on the edge of one fragment with those from other fragments to see which ones are a match. It can detect edges that seem to have eroded, and factor them in.

The team also has used a regular 2D flatbed scanner to capture the decorated front of each piece, but for now those images are used only for documentation, not matching. Eventually, the matching software will make use of both scans.

The project is one of many ways that archaeologists have gone high-tech. At the University of Pennsylvania, for example, David Gilman Romano and others have integrated satellite data with other measurements to map the layout of the Roman city of Corinth.

“There are lots of new archaeological methods that are based on scientific procedures that are changing the way archaeology is done,” Romano says. “And this is a good thing. This is creating new information.”

The excavations at Santorini are invaluable, says Philip Betancourt, a Temple University professor of art history and archaeology who is not involved with the Princeton project. That’s because few other sites from that era are so well preserved, having not been covered by volcanic ash.

The big eruption — believed to have occurred in 1630BC, though there is some disagreement — was preceded by several earthquakes. That apparently prompted the island’s Minoan inhabitants to flee, as no bodies have been found. It had been a thriving settlement, with ties to nearby Crete and possibly Egypt.

The first evidence of the ancient community was uncovered on Santorini in 1860, when workers were quarrying blocks to build the Suez Canal. Scholars began formal excavations in 1967, uncovering multi-storey buildings, pottery and the dazzling frescoes — what Betancourt calls one of the most important finds from the 20th century.

The paintings, which cover entire walls and even rooms in the mud-brick houses, depict landscapes, scenes of daily life and wildlife ranging from antelopes to monkeys.

So far, the Princeton computer scientists have done the bulk of the work with their new scanning technique. But they are headed back to the island in the next couple of weeks to drop off a laser scanner so the archaeologists can do it themselves.

“We really needed to have a system where somebody who knows nothing about computers can use this and scan thousands of fragments,” says lead author Benedict Brown, who recently left Princeton for a post-doctoral position in Belgium.

Though their stock-in-trade is technology, it turns out that even the computer scientists are able to use their eyes.

When the fake fresco was mailed to Princeton, one student who was not involved with the project happened to see the pieces spread out on a table, and decided to try assembling them by hand. He did pretty much the whole thing.

Vlachopoulos was in town then, and when he saw the results, he was surprised.

If computer science doesn’t work out, he told the novice puzzler, you can always get a job with us on Santorini.

MCT News Service

Decoding the past, Oman Tribune, October 3, 2008.