Premium
This is an archive article published on November 24, 2004

Computers put to work on detecting fake art

Is it a real Rembrandt or a fake? The wrong answer can cost a collector millions on the auction market, and leave a museum with a gaping hol...

.

Is it a real Rembrandt or a fake? The wrong answer can cost a collector millions on the auction market, and leave a museum with a gaping hole in its holdings.

A group of Dartmouth College researchers have put out a new technique they hope can help art specialists figure out how to avoid being snookered by an expert forger. The method, published in the National Academy of Sciences online report, is also sure to draw skepticism from many curators, who don’t believe their jobs can be reduced to a mathematical equation.

But with the art world still struggling to weed out fakes, Hany Farid, the study’s lead author, said the model he’s developing could help art historians who are open-minded enough to consider new authentication tools.

Story continues below this ad

‘‘What they’ve come up with here is a highfalutin’ pocket magnifying glass and that’s very valuable,’’ said Thomas Hoving, the former director of the Metropolitan Museum of Art and the author of False Impressions: The Hunt for Big-Time Art Fakes.

Farid, an associate professor of computer science at Dartmouth College, and co-authors Daniel Rockmore and Siwei Lyu have developed a method similar to that used to determine whether photos have been doctored or to analyse handwriting. After scanning a drawing or painting into the computer, they use a series of mathematical equations to determine the patterns distinctive to the artist. The model can be used to test the work against others credited to the artist.

If the data fall outside the style parameters for a particular artist, the work is clearly the work of another, the scholars contend.Henry Lie, director of the Harvard Art Museum’s Straus Conservation Center, said he would embrace the method if it worked.

‘‘It’s a big if,’’ Lie said. ‘‘It needs to be done again by other people and with other paintings and also reviewed from scientists who, unlike myself, work on that type of data to see if that science is valid. We have a long way to go but it’s certainly very interesting.’’

Story continues below this ad

Farid said he doesn’t expect his model to be the final answer on any art authentication questions. It will simply be another tool.‘‘There is no silver bullet,’’ he said.

Forgers have proved capable of fooling art collectors and museums for centuries. And authenticators have to deal with subtle questions, even when duplicity isn’t at play. For example, many of history’s most famous artists were also teachers. They might collaborate with students, or simply have their art willingly copied by an apprentice.

Farid used his method on a painting in Dartmouth’s collection credited to the Italian painter Perugino but generally believed to have been worked on by several different artists.There are six faces in the painting. Farid used a series of equations to create statistical measurements that could capture the style of the artist’s brushwork. Through that, he determined that at least four different artists contributed to the painting.

Farid acknowledges limitations to his method. It won’t work well with contemporary art, particularly those paintings that don’t have much contrast. The method won’t work well with an artist who has changed style, which can happen over years or even in the same painting. —NYT

Latest Comment
Post Comment
Read Comments
Advertisement
Advertisement
Advertisement
Advertisement