Today's digitization efforts lead to huge collections of scanned documents. However, the means for automatic preparation and further processing especially of ancient documents are still limited. In this paper, progress and implementation details of a framework for handling machine printed documents without traditional OCR-methods are shown. The approach is based on deriving any information needed for encoding directly from the original itself. This is achieved by extracting document specific alphabets and corresponding fonts. In particular, it is reported on how preprocessing, text segmentation, alphabet extraction, font generation, document encoding, as well as the repository work and interact. Moreover, the creation of ground truth data for evaluation and possible application scenarios for the system are shown.
Stefan Pletschacher, "Representation of Digitized Documents Using Document Specific Alphabets and Fonts" in Proc. IS&T Archiving 2008, 2008, pp 198 - 202, https://doi.org/10.2352/issn.2168-3204.2008.5.1.art00040