Art history in the making
Rice professor’s program brings Van Goghs back to basics

By MIKE WILLIAMS
Rice News staff

Vincent Van Gogh was picky about his canvas. That turns out to be a boon for art historians studying the master’s work using modern tools developed at Rice University.

	JEFF FITLOW
	Don Johnson, Rice’s J.S. Abercrombie Professor Emeritus of Electrical and Computer Engineering, shows a portion of a 44-painting mosaic of Van Goghs. On the left are the paintings, placed where they were cut from the original bolt of canvas. On the right are the matching weave patterns created from the X-rays by Johnson’s software, which allows paintings to be aligned edge to edge.

Fortunately, self-described “Van Gogh freak” Don Johnson, Rice’s J.S. Abercrombie Professor Emeritus of Electrical and Computer Engineering, is picky, too.

He and colleague Richard Johnson (no relation to Don), the Geoffrey S.M. Hedrick Senior Professor of Engineering at Cornell University, have developed a computer program to analyze aspects of Van Gogh and other paintings that can’t be seen by the naked eye.

The professors have come up with a signal-processing algorithm that automatically counts the thread density in an artist’s canvas from X-rays. The professors’ forensic investigations reveal previously unavailable details about the masters’ raw materials.

Until now, art historians have had to count threads by hand, a slow and tedious process that gives little information about how tightly a weave is packed. Don Johnson’s Fourier transform-based software analyzes everything at once, automatically creating a virtual fingerprint. That turns out to be useful for historians in attributing a painting or placing it within the timeline of a particular artist’s work.

Don Johnson is the principal investigator on a new grant from the National Science Foundation to take part in the Studio Practice Project at Amsterdam’s Van Gogh Museum. For a week each month next spring, he’ll join Richard Johnson, who will spend a semester at the museum building a database of Van Gogh’s work. Later, the team will apply the same techniques to the much smaller Delft School collection by Johannes Vermeer.

DON JOHNSON/RICE
Software analysis allows Rice’s Don Johnson to create a weave pattern of the vertical threads in an Old Master’s canvas. The stripped pattern is possible because X-rays cannot pass through lead in the white base coat of paint applied to raw canvas.

“They want to understand as much as they can about Van Gogh’s process,” said Don Johnson of the Van Gogh Museum’s motivation. “What kind of paints did he use? How did he paint?”

The museum, like many others, X-rayed artworks throughout the 20th century “almost as soon as X-rays were discovered,” he said. Historians were looking for details of the artists’ underlying paintings, for instance, or in Van Gogh’s case, for evidence of works he had painted over.

Johnson cares deeply about the art, but his real concern is what’s underneath. In that context, the painting is mere interference in the signal path. (In a paper published last year, the authors quaintly noted the “most important deviation from the ideal model is the painting itself, a phenomenon that is difficult to describe mathematically.”)

To see the canvas takes technology, some old and some new. Step one in the process is to acquire a digitized X-ray of an entire painting. Fortunately, museums have cabinets full of these. The Van Gogh Museum is no exception and is a terrific resource for Johnson and Johnson.

The image doesn’t reveal the canvas itself — the material, like human skin, is invisible to X-rays — but X-rays do not pass through the lead-white paint used as a base coat. The coat was commercially applied to bolts of linen artist canvas that could be two meters high and 100 meters long. Rolls were sold to dealers, who then sold the rolls or lengths cut from the rolls to artists.

Variations in X-ray intensity show where lead has pooled in the valleys of the weave, which tells the software what it needs to know about the threads. A spectral analysis of the X-ray reveals the number of threads per centimeter, vertically and horizontally, in a canvas. Vertical threads coming off the loom tend to march in lockstep, Johnson said, but the horizontals can vary in density, a critical factor that figures prominently in the fingerprint. The software also shows distortions in the weave created when a canvas was nailed to a frame for stretching.

Then the process gets interesting. With detailed thread-density and thread-angle-pattern maps of the paintings on his computer, Johnson uses a program to lay them out as though they were pieces of a puzzle. He then looks to line up the paintings in the order they would have been cut from the roll. Vertical threads may show very little variation, but horizontals reveal bar-code-like stripes that his software can use to match the paintings edge to edge.

“It’s hard to describe it,” Johnson admitted. “You really have to see it.”

And seeing is believing, in more ways than one. One in a series of weave-matched mosaics of dozens of Van Goghs – some famous, some not – reveals they all came from the same 100-meter bolt of canvas, cut into rolls. Those were purchased by Van Gogh’s brother, who kept Van Gogh stocked with supplies from Paris even during the painter’s years in an asylum.

“If you read Van Gogh’s letters, he was continually running out of canvas, begging his brother to send him more. He didn’t have any money. He was living off his brother — ‘Send me more canvas. I’m running out. I need it desperately.’

“Van Gogh was very precise: ‘I want this grade of canvas from this dealer.’ He really liked it, for some reason,” Johnson said. The mosaics make it easy to see that a particular series of paintings — the longest has 44 Van Goghs — came from the same bolt. But there’s no way to tell which subsets of those paintings came from the same roll.

“We think paintings that come from the same canvas roll, if we can reconstruct one, would really help in figuring out what was done when,” Johnson said.

The professor has pieced together 31 groups of paintings from the 300 or so he’s analyzed so far. “We think each corresponds to a different bolt of cloth,” he said. “The 44-painting one is the biggest. There are a couple in the 20 range, and whole bunch of pairs, which probably means they’re cut from a smaller piece of canvas. We’ve yet to put everything together. That’s what will happen next spring.”

When the Van Gogh project concludes next year, Johnson expects to give the Amsterdam museum a comprehensive archive of data about its collection that can be shared and compared with other museums.

He hopes to analyze more than half of the 864 Van Gogh paintings known to exist. Though Johnson doesn’t expect to get X-rays of the 200 or so paintings in private collections, he’s very close to getting half of the 600-odd museum-owned paintings. “I need about five more, and I’ll be over halfway on the available Van Goghs,” he said.

“Using weave matching to associate a painting to a roll or a bolt is still a hypothesis,” Johnson said. “We need a lot of other information to get the entire story together. That’s what will make next spring really exciting.”