The Counter-deception Blog

Examples of deceptions and descriptions of techniques to detect them. This Blog encourages the awareness of deception in daily life and discussion of practical means to spot probable deceptions. Send your examples of deception and counter-deception to colonel_stech@yahoo.com.

Sunday, April 03, 2005

 

Counter-Deception in the Arts

Who Really Wielded the Paintbrush?

December 23, 2004 NYTimes.com

By ANNE EISENBERG

Art historians have long used scientific tools to help them decide whether drawings and paintings are real or fakes, like counting isotopes in lead-based paints to spot anachronisms or shining X-ray and infrared radiation on oil portraits to discover what lies beneath.

Who Wielded the Paintbrush?

A statistical analysis of brush strokes done with a computer program suggests that four painters were involved. Without using a computer, however, one art specialist believes that the two saints on the far right and the one on the far left were not painted by Perugino.

Now researchers at Dartmouth College have introduced a forensic tool appropriate to the digital age: they have fed digitally scanned artworks into a computer, and then used image-processing techniques to create statistics describing the pen and brush strokes.

Like a connoisseur - a blend of Bernard Berenson and HAL - the computer analysis detected subtle differences in these strokes that might help distinguish an artist from an imitator.

First the scientists tried out their computational technique on drawings by Pieter Bruegel the Elder, the Flemish master, to see whether it could differentiate among landscapes accepted as Bruegel's and those considered bogus. Then they examined "Madonna With Child," an Italian Renaissance painting attributed to Perugino that depicts the pair with four saints. They hoped to determine whether one set of hands created the work, or as some art historians think, Perugino assigned parts of it to his students.

The computer program agreed with the Bruegel experts, grouping eight drawings attributed to Bruegel in one category and assigning the imitations to a separate pile. The analysis of the brush strokes in the Perugino suggested that four sets of hands contributed, jibing with the view of some art historians but not others.

David Donoho, a statistician and professor at Stanford University, said that although the software did reasonably well in its debut performance, researchers had far to go. For one thing, he said, the method was used only on a handful of artworks. "It's a first step, but a promising one," he said.

Yet some art historians are quick to point out that the technique disregards factors like an artist's technical skill.

"The computer was asked to analyze only a few aspects of the painting," said Laurence Kanter, a Renaissance art specialist who is the curator in charge of the Robert Lehman Collection at the Metropolitan Museum of Art. "And it missed fundamental ones like how color was applied and mixed."

Dr. Kanter has his own opinions of the Perugino based on the old-fashioned method - looking at it. On a visit to the Hood Museum of Art at Dartmouth, which owns the painting, he decided that Perugino in fact painted three of the six figures. As for the rest, he said, "The two figures on the right are by someone mechanical and second-rate, and the figure on the far left is by someone more capable and technically proficient than Perugino."

The computer did not register these details, he said. "That's why these programs are potentially interesting, but need to be developed," he added.

An article by the researchers ran this month in the Proceedings of the National Academy of Sciences. One of its authors, Daniel Rockmore, a mathematician and professor at Dartmouth, said the project began when he toured a Bruegel show in 2001 at the Metropolitan Museum of Art with the exhibition curator, Nadine Orenstein, a Bruegel expert at the museum. The show gathered actual and imitation Bruegels in one space. As Dr. Rockmore looked at the landscapes, he began to wonder if a computer could do what experts had done, identifying the real drawings.

"There's a long history of drawings attributed to Bruegel that we now know are not by him," he said. It occurred to him, he said, that "maybe the mathematics of modern digital-image processing could see it, too," particularly if he used the mathematics of wavelets.

"This type of mathematics is good at determining the characteristics of lines and curves, or jagged lines and jagged curves," he said. Such an analysis, he thought, might be perfect for distinguishing Bruegel's characteristic lines and shadings.

To perform the analysis, Dr. Rockmore collaborated with a colleague and co-author of the paper, Hany Farid, an associate professor in the computer science department at Dartmouth, who has long experience in characterizing images using wavelet-based statistics. In particular, Dr. Farid had been working on computational algorithms to determine when a digital image had been doctored.

"I was already in the business of thinking about the statistical properties of an image," Dr. Farid said.

Dr. Orenstein lent the group her slides of the Bruegel show, and the team, joined by a graduate student, Siwei Lyu, set to work. In the resulting data, Dr. Rockmore said, "all the points representing the true Bruegels clustered together," and all the non-Bruegels were outside the cluster.

"It's certainly not definitive," he said. "But it's amazing that mathematics could distill differences that the experts can spot by eye."

The researchers then turned their attention to the Perugino. "Right now forensic science can't answer the question of how many hands may actually have done a painting," Dr. Farid said. "But if you are buying a Raphael at auction, you might want to know what percentage he actually painted."

To see if their algorithms could detect different characteristics in the strokes of the artists' brushes, the group started by using a large-format camera to produce an 8-by-10-inch negative that they then scanned, converting it to a digital image for processing.

They extracted the six faces in the painting - those of the Madonna, baby Jesus, and the saints Anthony Abbot, Francis, James and Benedict. Using the same method they did with the Bruegel, they looked for consistencies and inconsistencies in characteristics like the texture of the strokes. The analysis, however, was not done in full color. "We converted it to gray scale first because we didn't want to pick up on simple color differences between faces," Dr. Farid said.

Dr. Kanter, the Lehman curator at the Met, said that omission was one of many weaknesses in the technique that would need to be addressed.

Far larger samples will be needed before art historians can take the work seriously, said Dr. Orenstein, the Bruegel scholar at the Met. "The computer would need to draw on a huge database for each artist," she said. "It's going to be decades before they have a connoisseur meter and go around to auction houses."

Dr. Orenstein and Dr. Kanter said that Rembrandt would be a good candidate should researchers decide to assemble data on a large body of work. "There is so much discussion of his work, and so many students," Dr. Orenstein noted.

Conservators have also expressed interest in the method. At the Williamstown Art Conservation Center in Massachusetts, Kate Duffy, a chemist and head of the analytical services department, plans to meet with Dr. Rockmore in January. "It's an exciting tool," she said of the computer technique, one that may someday gain a place in her laboratory already full of chromatographic, X-ray, and other equipment. "We are going to try some things out with it."

Ellen Handy, chairwoman of the art department at the City College of New York, also said the tool seemed promising. "It may help us historians answer some of the most basic questions in the field, like 'How many people made this piece of work?' "

Dr. Handy, who once led a master's degree program in connoisseurship, said the software might buttress the insights of mere humans. "It's a new tool to back up the intuitive judgments of smart people with good eyes," she said.

Dr. Donoho said that he wondered why it took so long to develop statistical tools that could help curators authenticate artworks. "Statistical pattern recognition is useful," he said. "We see it all the time, for example, to rate the risk of credit transactions." Spam filters are also based on statistical decision theory.

"It's about time statistical tools were applied in an artistic context," he said.

Comments:
sell off vacation
 
Post a Comment

<< Home

Archives

September 2004   October 2004   November 2004   December 2004   February 2005   April 2005   July 2005   August 2005   September 2005   October 2005   November 2005   December 2005   January 2006   February 2006   March 2011   June 2011   August 2011   September 2011   May 2012   February 2017   June 2019   August 2020  

This page is powered by Blogger. Isn't yours?