Past Awardee

Cyber Connoisseurship: Tools to Aid Understanding of the Medieval French Book Trade

Anne D. Hedeman

College: Fine and Applied Arts
Award year: 2008-2009

I share an interest with Peter Ainsworth (Department of French, University of Sheffield, UK) in a corpus of illuminated manuscripts of Froissart's Chronicles that he has successfully digitalized and mounted on the web (viewable in Virtual Vellum, described more fully below) and that is being studied by a team of international scholars. The high resolution digital scans of these manuscripts are stored at various sites through member institutions in the Worldwide Universities Network (WUN); UIUC currently houses the high resolution digital files for one Froissart manuscript on an Atlas server in the College of LAS. The goal of my proposal to become an NCSA Faculty Fellow is to work with Peter Bajcsy to develop cyber tools for analyzing the visual imagery embedded in these Froissart manuscripts, in order to provide insight into both the construction of these specific books and, more broadly, the functioning of the medieval Parisian book trade, which we will make available on a website shared by NCSA and Medieval Studies at UIUC and with the University of Sheffield in the UK.

Our work will complement the Engineering and Physical Sciences Research Council (EPSRC, UK) funded "Pegasus" project led from the University of Sheffield and its Humanities Research Institute by Peter Ainsworth and Michael Meredith. The resources offered by our English colleagues consist of UCSD's Storage Resource Broker (SRB) clientware used on the WUN Grid node, and "Virtual Vellum" an online manuscript viewing and manipulation tool, funded by the UK's Arts and Humanities and Engineering and Physical Sciences experimental e-Science program. Our contribution will be the development of a toolkit that enhances and complements these resources to create a robust cyberenvironment that supports the study of virtual manuscripts and other high-resolution images.

To analyze artistic production, we propose a methodology utilizing machine learning and pattern recognition techniques. In order to compare multiple manuscripts and different artistic hands at work, we will first identify a set of original image scans for learning. Image scans will have to be selected manually by Hedeman and a graduate student in Art History since very little is known about these artists. Since image scans contain both illustrations and text, Peter Bajcsy and a graduate student in computer science will address the problem of automatic cropping in order to carve out image sub-areas that contain only illustrations. This problem will be approached by image scan tiling and histogram-based classification of text and illustration regions. To this end, we will leverage algorithms originally designed for classification of Lincoln's writings, where text regions are characterized by document-specific "but unique" combinations of ink and paper colors. We will research similar approaches to automatically segment illustrations from the image scans of multiple manuscripts.

Both the tools that we develop and the results of our analysis will be made available to researchers, educators and students through NCSA and Medieval Studies websites at UIUC and through the Worldwide Universities Network (WUN) and "Virtual Vellum."