In this paper we present some recent results in using a multi-scale,
multi-orientation decomposition analysis (e.g., wavelets) of high
resolution digitized versions of drawings and paintings as an aid to
authentication. To date, most techniques for authentication in the
arena of paintings and drawings have relied on some mixture of
physical forensics (e.g., material dating), historical documentation
(i.e., the work's provenance), as well as the opinion of the
connoisseur, a person with a lifetime of experience in viewing works
by the artist in question. Our preliminary results indicate that
various statistics acquired from the wavelet decomposition of a work
seem to provide a "digital signature" of the artist, in the sense
that the statistics of works of a given class by the same artist
cluster together, while remaining apart from imitations.
|
Related material: pnas04
|
Home Publications Research |