Stork’s Computer Vision Revolutionizes the Study of Art

Reported by Catherine Yochum ’15

David Stork (image courtesy of http://www.diatrope.com/stork)

David Stork (image courtesy of http://www.diatrope.com/stork)

In 1430, something changed in Renaissance art: suddenly, the flat, stilted figures portrayed by painters such as Giotto and Masolino were replaced by contoured, humanistic figures and faces displaying near-photorealistic emotion. Was it some kind of breakthrough in optical devices that triggered this shift? David Stork, whose recent lecture at Bowdoin was sponsored by the Digital and Computational Studies Initiative, uses computer image analysis to solve this and other enigmas that have previously eluded art historians.

Stork – a scientist and scholar who has held faculty positions in physics, math, computer science, art and art history, electrical engineering, statistics, neuroscience, and psychology - is research director at the technology company Rambus Labs. He has presented more than 200 courses, lectures, seminars, and colloquia on computer image analysis of fine art and authored the book Seeing the light: Optics in nature, photography, color, vision and holography.

To get an idea of how Stork analyzes famous paintings, take Johannes Vermeer’s Girl with a Pearl Earring, a work considered to be a character study on youth and beauty rather than a specific portrait. Art historians have long wondered whether Vermeer painted from a live model. To solve this art history mystery, Stork took advantage of the painting’s the soft, glowing light. He modeled a reflection on a glassy pearl earring until it fit what he saw in the painting. He created a computerized eye with a catchlight that matched the girl’s, taking into account the eye’s natural corneal bump and the direction of the subject’s gaze. He measured the direction of the light washing over the occluding contour of the anonymous girl’s face. He even made a rotatable digital model of the figure as a whole. He then analyzed the angles from which each part of the painting was illuminated.

Courtesy of David Stork

Courtesy of David Stork

The consistency in the angles was uncanny, suggesting that Vermeer must have used a live model: it would have been nearly impossible for these light sources to match so closely if he had painted from his imagination. When Stork presented his findings to art conservators at The Hague’s Mauritshuis Museum, the painting’s current home, “they just went crazy” with excitement over the revelation.

The same kinds of light analyses can extend to tell us whether parts of an image were finished at a later date or time, or by the hands of an atelier rather than a master. In Georges de la Tour’s candlelit Christ in the Carpenter’s Shop, a technique called cast-shadow analysis shows that the artist truly did paint by candlelight. Complementing Stork’s analyses of light are the 3D imaging techniques that he tailors to specific paintings. For example, his computer models have digitally broken down Van Gogh’s Self-portrait with Gray Felt Hat layer by layer to see “what it looked like to Van Gogh” as the artist constructed the image on top of its abstract, blue base.

So what has Stork learned about that mysterious change in 1430, when painters started capturing realistic human figures and faces? The shift was not, in fact, a result of new optical technology, as some art historians have hypothesized. In Jan van Eyck’s Arnolfini Portrait, for example, Stork digitally analyzed the angles and symmetry of the painting’s prominent gold chandelier, showing that it would have been impossible for these unrealistic proportions to have emerged from an optically traced image. That leaves plenty of other factors that could have been responsible, such as the rise of slow-drying oil paints, trends in realist sculpture, and a new emphasis on humanism.

Not only are Stork’s insights changing what we know about the history of art, but they’re making a difference in how we treat treat paintings today. Using computer imaging and geometric principles, Dr. Stork has recently demonstrated that Self-portrait in a Convex Mirror, a late renaissance painting by Parmigianino, takes on a new and apparently more correct perspective when hung at a lower height. The viewer’s line of vision, Stork argued, should align with where Parmigianino would have seen his own eye – that is, in the center of the convex plane. The Kunsthistorisches Museum has re-hung the painting since Stork argued this case in Vienna in 2009, and he is currently working to publicize photographs of the painting “as it should be seen.”

Given that his work is revolutionizing our understanding of famous works by master artists, does Stork hope that computer image analysis will shape popular perception of who was “better” at art? “Absolutely positively not,” he says. A taste for fine art should still be left to the human eye.

thumb:"River St. 2., Lewiston, ME" (Michael Kolster, 2010)