Tuesday, December 28, 2004

Math Could Help Art History

Authors of novels have been identified through context-free word counts. Similarly, one might be able to identify painters by analyzing the frequency of certain types of curves. This was the idea of the mathematician Hany Farid and two colleagues at Dartmouth College in Hanover, New Hampshire, U.S.A. [1].

They have developed an approach that builds a statistical model of an artist from the scans of a set of authenticated works against which new works are compared. The statistical model consists of first- and higher-order "wavelet decomposition." They have analyzed 13 (8 true and 5 false) drawings that have been attributed to the 16th century artist Pieter Bruegel the Elder, and have gotten the results that confirm expert authentications [2].

The researchers have also applied these techniques to determining the number of artists that may have contributed to the painting "Madonna with Child" attributed to the 16th century Italian painter Pietro Perugino, and again have achieved an analysis agreeing with expert opinion [2].

This indicates the possibility that the computer can help research in art history, rendering one of examples of cooperation between arts and science in a broad sense.
  1. "Verifying art with math" Science, Vol. 306, 1678 (2004).
  2. S. Lyu, D. Rockmore and H. Farid, Proc. Natl. Acad. Sci. Vol. 101, 17006 (2004).

No comments: