Accessibility links

Computers Help Authenticate Paintings

Art collectors spend millions to buy masterpieces, but the masterpiece they buy today could prove to be virtually worthless tomorrow if they aren't careful. Now, researchers at Dartmouth University say they have developed a technique that uses the power of a computer to help collectors determine whether the investment they've made is a sound one.

The technique begins with a digital photograph of a painting, drawing or print that captures the subtle texture and brushstrokes of the artist.

"It's very, very high resolution. Tens of thousands of pixels by tens of thousands of pixels," says Hany Farid who heads the team that did the study at Dartmouth. He says the computer then analyzes the image in the same way similar programs have analyzed literature to determine authorship.

"When people were trying to authenticate Shakespeare's sonnets, they looked at a previous body of work believed to belong to him and they compared statistically that body with the works in question," he explains. "The statistical measurements there are word distribution, length of sentences, punctuation, things of that sort. We do a similar thing. It's just that the statistics are now in respect to images."

Mr. Farid says the technique uses the same type of technology that compresses the digital information in a photograph so it takes up less memory. He refers to a study done using a painting attributed to Italian Renaissance artist Pietro Perugino, which experts believe was completed with help from some of his students.

"Let's consider the master and the way he may paint," he adds. "Probably very smooth brushstrokes, very consistent, very delicate, very elegant lines. And when the imitators try to imitate, probably a little bit jerky, because we know that when you try to mimic somebody's actions it never comes out smooth. Now let's imagine compressing the image of the master. Probably we'll have a lot of redundancy because of the consistency in which it's painted. Whereas the student is less consistent, so there are less redundancies and it will not compress into as small a file."

Since different portions of the painting compressed to different-sized files, Hany Farid and his team of Dartmouth researchers determined that different artists painted them.

Of course, artists are not always consistent. They have been known to radically change styles. Take the 19th-century master, Vincent Van Gogh, for example, whose brushwork changed dramatically from his early, dark Dutch paintings, to the colorful landscapes he did in his final years in southern France. Hany Farid says the technique he and his team developed to analyze art is only good within a narrow time period. He emphasizes that it is only a tool, one that still requires an expert's reading of the data.

"This does not tell you who did what," he explains. "It simply tells you about consistency. We say we believe there are many hands present in this painting. And the art historian may come in and say, 'This figure on the left. This is the best quality, so this probably belonged to the master.'"

In addition to analyzing the Perugino painting, Mr. Farid says the technology was recently applied to thirteen drawings to determine which were done by Flemish artist Pieter Bruegel and which were imitations. Their conclusions were consistent with those of art historians. The Dartmouth team plans to do more tests over the next few years, with hopes that the technology will eventually become a standard authentication tool for the worldwide art collection community.