Using high resolution digital images of drawings by Bruegel and some of his imitators, as well as a painting by Perugino, the computer scientists captured data about pen or pencil stroke patterns and other elements that represent an artist's style or aesthetic signature. This signature was then used to discover consistencies and inconsistencies within a single piece of artwork or among works by the same artist. The study was published in the Proceedings of the National Academy of Sciences (PNAS) Online Early Edition during the week of Nov. 22-26, 2004.
"Similar methods have been used to analyze works of literature, like assigning authorship of different parts of The Federalist Papers to Alexander Hamilton and James Madison," says Hany Farid, Associate Professor of Computer Science and an author on the paper. "We can find things in art work that are unique to the artist, like the subtle choice of words or phrasing and cadence that are characteristic of a certain writer."
Farid, with Siwei Lyu, a Ph.D. student, and Daniel Rockmore, a professor of Mathematics and Computer Science, tested their technique on one painting by Perugino thought to be the work of multiple hands, and another 13 works that had been at some point attributed to Pieter Bruegel the Elder. Their results confirm what the art experts and art historians have concluded. The Madonna with Child by Perugino shows evidence of at least four artists, and of the 13 drawings, only eight are authentic Bruegels.
"We've been able to mathematically capture certain subtle characteristics of an artist that are not necessarily visible to the human eye. We expect this technique, in collaboration with existing physical authentication, to play an important role in the field of art authentication," says Farid.
Farid's research is funded by an Alfred P. Sloan Fellowship and the National Science Foundation. Rockmore's work is supported by the National Science Foundation and the National Institutes of Health.