Previous methods for writer and handwriting identification have based their analyzes primarily on documents that were written in a specific language or with defined graphic symbols and that are not damaged in any way.
In cooperation with the Humboldt University of Berlin and the Fraunhofer IPK, the MusterFabrik Berlin is currently working on also identifying writers and the handwriting in historical documents with age-related damage, yellowed pages and with faded or otherwise worn ink. For that purpose, a context-specific structural analysis of image patterns is conducted by means of machine learning algorithms (deep learning, neural networks).
The high-quality digital copies of the 18th century manuscripts from Böhmisch-Rixdorf in the Czech language produced using the digitization technologies of the MusterFabrik Berlin serve as the basis for the application of the image processing algorithms.
The project is funded by the Volkswagen Foundation.