Extracting the Hidden: Paper Watermark Location and Identification
Roger Boyle & Kia Ng (Leeds)
Digital Classicist/ICS Work in Progress Seminar, Summer 2009
Friday 3rd July at 16:30, in room STB3/6, Senate House, Malet Street, London WC1E 7HU
We consider the recovery and identification of watermarks from archaic documents.
Watermark studies go back many years and have proved fruitful in document dating and identification. Many techniques have been developed but the advent of large digital repositories and advances in imaging algorithms presents new opportunities.
We have conducted two attacks on these problems: 1. A "bottom up" pixel-first approach that, while similar to some recent literature, seeks to automate parameter selection to be more applicable and robust. 2. A "top-down" modelling approach that has shown success in very challenging materials and succeeded in identifying details of marks that had not been seen before.
We have chosen a back-lighting approach: this delivers good quality, digitally-native images with minimal risk to the original documents in a manner that is relatively portable and cheap. We have exhibited our approaches on a wide range of documents, notably some of the Arabic holdings in the University of Leeds that have attracted significant scholarly interest from Arabic codicologists, and which represent very difficult imaging challenges.
Typically the retrieved watermark designs from individual pages are blurred and not precise. For better definition of the extracted design we align and superimpose multiple watermark patterns, combining them to offer a sharper and more informative result.
We have further looked at the possibility of using our approach to distinguishing individuals from twins (the common mode of paper milling being in pairs). We also consider we can contribute to other mould identification tasks, such as characterising chain lines: historians of paper often find these as (or more) useful.
The seminar will be followed by wine and refreshments.