Paintings that have lain hidden for hundreds of years underneath some of Europe’s most treasured artwork are being revealed with remarkable new clarity with the aid of 21st-century technology and signal processing experts at CU-Boulder.
Assistant professor of electrical engineering Shannon Hughes has been collaborating with the University of Antwerp, Belgium; Delft University of Technology in The Netherlands; and colleagues at several other U.S. schools on the virtual recovery and restoration of long-concealed paintings.
Famous works by the likes of Van Gogh, Rembrandt, and Runge were often painted on top of other paintings that are just now being detected with the advent of X-ray fluorescence imaging technology.
As many as 20 paintings, or about 15 percent of those recently X-rayed at the Van Gogh Museum in Amsterdam, were found to have hidden paintings beneath them, according to Hughes, who interfaces regularly with art conservators, historians, and others in the fine arts world.
The discovered works represent a new area for art historians that could shed light on issues of authorship and detection of forgery, as well as provide a new set of paintings by renowned artists to be studied and enjoyed.
X-ray fluorescence imaging is a noninvasive technique that involves measuring the concentrations of chemical elements at different spatial locations across a canvas as an indicator of paint pigment.
When combined with advanced signal processing tools being developed by Hughes and her colleagues, it allows for hidden images to be revealed in all their glory without any damage to the surface painting.
Hughes became involved in the field a few years ago when she was working on her PhD in electrical engineering at Princeton University. Chemistry professor Koen Janssens of Antwerp, along with materials expert and art historian Joris Dik of Delft, were looking for a collaborator to help clean up and restore data they were obtaining through X-ray fluorescence.
“They are the best in the world at creating high-quality chemical imaging of fine art, but they needed help to clean up the chemical images and integrate them together into a single color image. So we have applied some of our signal processing tools and developed some new ones as well,” says Hughes.
Hughes’ team has developed several new methods to repair artifacts introduced through the imaging process, remove unwanted features that sometimes result from heavy use of paint on the surface paintings, and identify and repair small areas of information loss.
One of the images that she and her students have worked on—and improved the quality of considerably—is a portrait of a peasant woman that lies under Van Gogh’s 1887 painting Patch of Grass. The portrait was painted in shades of brown and red, whereas the surface painting is dominated by hues of green and blue, allowing for a relatively clean separation of chemical elements.
While previous research had discovered a vague outline of a head behind the painting, it was not revealed in great detail until 2008. A paper published this spring in the journal Signal Processing, with recent CU master’s graduate Anila Anitha as lead author, reveals an improved image and details the novel processing techniques used by the research team.
Another reconstructed image the CU team has studied is a partially obscured painting beneath a Philipp Runge portrait of the German Romantic period called Pauline in a White Dress against a Summer Landscape. The hidden painting is believed to be a more intimate portrait of the painter’s wife that was painted over to give her a more conservative appearance.
Hughes says that the team can use similar techniques for distinguishing the brushwork of one artist from others for the detection of forgery. This is a daunting task, as even art experts disagree on the attribution of some paintings, but quantitative analysis could provide additional evidence to help settle these disagreements.
Her team has attempted to address the task by using a combination of filtering, statistical modeling, and machine learning. “In recent experiments, we have found that we can distinguish the brushwork of Vincent van Gogh from that of other artists with 92 percent accuracy on a test set of over 100 paintings,” Hughes says.
Ilana Trumble, a junior majoring in applied mathematics, is working on reconstructing the hidden Runge portrait with support from the college’s Discovery Learning Apprenticeship Program. Her research builds on previous color estimation work that has been done on the painting, using data analysis and programming tools.
“It’s a really cool way of linking math and computing with history,” says Trumble, who selected the project from a long list of research opportunities offered by faculty last spring.
She and nearly four dozen other students in the Discovery Learning program will present their posters at a research symposium on April 19.