Image Analysis for Dating of Old Manuscripts

This paper presents an application of digital image processing to historical sciences. A major tool for dating old undated documents are watermarks found in the paper. Hardcopies of the watermarks are scanned, preprocessed, improved and contrast enhanced by adaptive digital filtering methods for printing, storing in an image database, and extracting the watermark as a set of strokes from the image. For extraction, a semiautomatical procedure is suggested. The extraction result is a short sequence of cubic spline curves representing the watermark fully and allowing to select identical or similar watermarks from the existing database.