Looking at art through science
David Stork of Portola Valley develops computer imaging methods to look beyond the paint of great artwork
It's lunchtime at Quadrus Cafe, a short distance from scientist David Stork's workplace on Sand Hill Road. Tables in the naturally lit dining room on this late spring day are covered simply with white paper that might have been chosen so as to not compete with the striking modern paintings on the walls.
Mr. Stork arrives a few minutes late for lunch and an interview with the Almanac, arranged so that he can talk about the groundbreaking work he's doing in an area that straddles the often-alienated worlds of art and science.
The conversation moves energetically, and by the end of the meal, the stark white paper beneath the soup bowls and drinking glasses has been transformed into a sketchbook of ideas, explanations and emphasis. There are images of optic lenses that might eventually find their place in a more efficient digital camera. Swirling, flat-noodle lines suggestive of Van Gogh's brushstrokes in a famous self-portrait.
There's a rectangle with two horizontal lines and several vertical slashes — in the style of Piet Mondrian, it's an example of an image that naturally satisfies the eye. If Mondrian had put this vertical line over here rather than there, Mr. Stork says, planting his unrelenting silver pen in the center of the rectangle, the painting wouldn't work.
Ultimately, the impromptu sketchbook represents a number of key aspects of the Portola Valley resident's life: his work as chief scientist at Ricoh Innovations (that's where the optic lenses come in); his intense enthusiasm for the fine arts, and how meaning is conveyed by images; and his independent scientific work in developing computer methods to uncover various secrets of great artists' work.
A recognized leader in the field of pattern recognition and computer visioning who holds a doctorate in physics, Mr. Stork spends his evenings, weekends and, sometimes, vacation time outside the office on his arts-related pursuits. He and several collaborators have developed computer methods of analyzing light, patterns and textures in paintings that he hopes to see applied widely to the study of art history and interpretation.
He has used those methods to examine nuances of paintings by Caravaggio, Vermeer, Velazquez and other esteemed masters of centuries past.
His examination of Van Gogh's "Self-Portrait in a Grey Felt Hat" uses a method first developed with colleagues in Stanford's Department of Statistics, where he is a consulting professor, to digitally "peel away" seven layers of brushstrokes from the work to ascertain the artist's approach to laying down the paint.
"Wouldn't it be good to be able to see Van Gogh paint?" he asks, looking up from his tabletop sketch. "Well, this is the best we can do."
Although Mr. Stork's interest in art was sparked when he was a child, growing up with a photographer father and a grandfather who collected works by major American painters, the multi-discipline scientist's entrance onto the world stage of art history and appreciation was not exactly planned.
The grand entrance
In 2001, prominent American artist David Hockney published "Secret Knowledge," a book pushing his claim that Renaissance painters used lenses and mirrors to project images onto canvases and other surfaces that would then be traced by the artist. It was that technique, he argued, that explained why paintings after around 1425 took on characteristics of three-dimensionality and depth.
Although Mr. Stork hadn't been paying much attention to the resulting debate among art historians, he had to get up to speed on the issue fast when he was invited to a conference to discuss the question from a scientist's perspective. "I found the theory fascinating, and there was maybe a chance that it was right," Mr. Stork recalls.
With only two weeks to prepare, he launched his own tests on the theory, and concluded it had no merit — a conclusion, he says, that is now widely accepted in the art world.
The experience turned out to be a launchpad for an exciting new application of his range of scientific expertise. Since his work refuting the Hockney theory, Mr. Stork has published a large number of papers on computer analysis of art, and given more than 230 lectures in 18 countries, he says.
Bridging the gap
One of his latest lectures, given in April at Ithaca College, was "When Computers Look at Art: Image Analysis in Humanistic Studies of the Visual Arts." The talk was part of the college's C.P. Snow Lecture Series, and Mr. Stork was unquestionably an apt choice to present it: The series is named after the British novelist and physicist whose lecture more than 50 years ago, "The Two Cultures," published later in essay form, argued forcefully for the need to bridge the gap between science and the humanities for the overall benefit of society.
Mr. Stork says he spoke in that lecture of "the complementary way ... science can be used to bring art history back to the careful looking at objects of art." Promoting a "careful looking" approach to art appreciation, which he refers to as connoisseurship, is a key objective of his computer analysis of painting. He says he is currently working on a paper with art historian David Stone focusing on how new techniques in digital analysis "will reinvigorate connoisseurship" in painting.
After a period of several decades, he says, "theory swept humanities departments," emphasizing political and philosophical concerns in art history. "Recently, there seems to be a return to art as objects, I'm happy to say, and I and a number of enlightened art scholars believe my computer methods may help in this return to close visual study of art."
Art historian Bryan Wolf, a Stanford professor and co-director of the Stanford Arts Initiative, says Mr. Stork's work in art analysis is "valuable and important" because it can help art historians assess how a painting was done — "the history of a painting as it emerged on the canvas."
Art historians are likely to benefit the most from digital analysis of paintings because, with the revelation of nuances of the artwork's progression and the artist's approach, a scholar is led "very quickly to much larger historical questions," including questions about the culture the artist was a part of.
But does the work of Mr. Stork and others delving into the technical aspects of painting matter to the lay person? "Yes, but not because you're a technician or a scholar — that's not what you need to care about," Mr. Wolf says.
"The reason you would care is not from a connoisseur's view, but because (the revelations that result from the work) help us understand history better, and art's history better." And that, he adds, "helps in being a better citizen of the world."
Mr. Wolf notes that Mr. Stork's work is not the only method available to closely examine paintings. "But he adds to the arsenal of tools available to answer the questions" art historians try to answer.
He also adds a word of caution: "When you come up with scientific or empirical computer-based methods of re-examining art, there is a danger that instead of seeing it as a tool to take you somewhere else, you'll say ... that it explains the art." But the examination can't stop with what can be gleaned from the science, he says, because that would make the discovery "interesting like a piece of gossip, but it doesn't teach me anything." Scientific scrutiny, he adds, "needs to be used in the service of teaching something."
Mr. Stork says his work is intended as a tool to help art historians rather than an end in itself. "All my work is focused on answering questions the art community cares about."
In addition to his work at Stanford, which includes teaching a course on image analysis in art, Mr. Stork has had faculty appointments at numerous universities in disciplines including physics, mathematics, computer science, statistics, electrical engineering, neuroscience and psychology, he says.
With all his interest in great art and painting, does he paint? A little, he says. "I'm good enough to know how bad I am." But the pursuit helps him understand the painting process better.
He now has a few projects at some stage of development, including computer image analysis in the authentication of Jackson Pollock's drip paintings, he says.
Although his analysis of individual paintings occupies much of his time outside his work at Ricoh Innovations, a bigger question involving art and imagery occupies his mind: How do images convey meaning?
It's "one of the deepest problems in all of science, and the humanities," he says.
Will he be the one to answer the question? He smiles cautiously over his sketches and iced tea. "I'm working on it," he says, but in little bits. It's unlikely he'll give up, though. "Once you get that question — that bee in your bonnet — you have to work on it."