The Archimedes Palimpsest is a tenth century manuscript that contains the earliest known copies of the treatises of Archimedes, two of which were previously uncovered. Several multispectral imaging techniques have been applied to high-resolution images of the palimpsest to provide scholars of ancient Greek mathematics with visible text for inquiry. It is estimated that nearly twenty percent of the Archimedes Palimpsest project is currently undergoing an experimental imaging phase to assist in this final translation effort. A recently developed character recognition tool has been moderately successful on the degraded regions of the text. The goal of this research is to develop a method for integrating contextual information into the existing character recognition system, based primarily on correlating the spatial properties of image data. A simple case study was first performed to survey relevant pattern classification tools. Bayesian networks represent an efficient means for describing the uncertainty in a given system. The proposed network, designed specifically for the Archimedes Palimpsest, is discussed in detail. A preliminary version of this Bayesian network has been developed, and results are presented.
Mr. Walvoord is a doctoral candidate from the Carlson Center for Imaging Science at RIT. Part of this research was completed as an independent study project under the direction of Dr. Roxanne Canosa, Assistant Professor of Computer Science at RIT.
Colloquia Series page.