An OCR System for Printed Documents

This paper describes the general structure of a full automated document analysis system for printed documents. The system is based on a character preclassification stage which reduces the number of patterns to recognize and introduces a new contextual processing. This specific approach for multifont printed documents reading is based on pattern character redundancies. With the study of prototype pattern instead of characters, we can extract reliable contextual information on the entire text. This statistical information is used to check hypotheses given by the multifont recognition stage. We shall describe the system organization and particularly the preclassification stage, then we shall give some results on recent documents.

• Type: Conference
• Year: 1992
• Where: MVA
• Authors: Frank Lebourgeois, Jean-Luc Henry, Hubert Emptoz
• Keywords: Character Preclassification Stage, Multifont Printed Documents, MVA 1992, MVA 2007, Preclassification Stage