Document reconstruction by layout analysis of snippets

Abstract

Document analysis is done to analyze entire forms (e.g.intelligent form analysis, table detection) or to describethe layout/structure of a document. Also skew detection ofscanned documents is performed to support OCR algorithmsthat are sensitive to skew. In this paper document analysisis applied to snippets of torn documents to calculatefeatures for the reconstruction. Documents can either bedestroyed by the intention to make the printed contentunavailable (e.g. tax fraud investigation, business crime)or due to time induced degeneration of ancient documents(e.g. bad storage conditions). Current reconstructionmethods for manually torn documents deal with the shape,inpainting and texture synthesis techniques. In this paperthe possibility of document analysis techniques of snippetsto support the matching algorithm by considering additionalfeatures are shown. This implies a rotational analysis, acolor analysis and a line detection. As a future work it isplanned to extend the feature set with the paper type(blank, checked, lined), the type of the writing(handwritten vs. machine printed) and the text layout of asnippet (text size, line spacing). Preliminary results showthat these pre-processing steps can be performed reliablyon a real dataset consisting of 690 snippets.