Invited Talk: Robert Sablatnig (TU Wien) – Multispectral Imaging and Writer Identification for Historical Manuscripts, 20th of January, 12h noon
It’s a great pleasure to welcome Prof. Dr. Robert Sablatnig to our lab to present his latest research!
Time: 12h noon
Title: Multispectral Imaging and Writer Identification for Historical Manuscripts
Abstract: The Computer Vision Lab @ TU Wien is working on cultural heritage related fields for more than 20 years. This presentation gives an insight on the latest achievements in the area of multispectral imaging as a prerequisite for analyzing historic manuscripts and on Automatic Writer Identification (AWI) as one of the analysis fields in the area of historic documents. MultiSpectral Imaging (MSI) has become a popular tool to reveal properties and structures in cultural heritage objects that are hidden to the human observer. One of the inherent problems of MSI applications is chromatic aberration. Due to an extended spectral range, the effect appears more pronounced than in conventional photography in the visible spectrum. Our recent work is concerned with longitudinal chromatic aberrations, i.e. shifts of the focal plane along the principal axis of the camera, as they are hard to correct in post-processing and should be avoided during acquisition. To this end, a calibration scheme to measure the wavelength- and distance-dependent focal shift behavior of a given camera/lens system is proposed, which allows for a mechanical compensation at acquisition time. The images taken are the basis for the subsequent AWI task, which has received a lot of attention in the document analysis community. However, most research has been conducted on contemporary benchmark sets. These datasets typically do not contain any noise or artefacts caused by the conversion methodology. Therefore, current state-of-the-art methods in writer identification perform differently on historical documents. In contrast to contemporary documents, historical data often contain artefacts such as holes, rips, or water stains which make reliable identification error-prone.