Reconnaissance dynamique des formules mathématiques

Abstract

This dissertation is about the development of a program, attached, about optical character recognition, especially adapted to the treatment of mathematical formulae. The main difficulty compared with usal OCR is the following one: mathematical formulae have a complex organization, and their reading can't be done in only one direction. This program is an improvement to David Léonet's one, based on the works of Jérôme Anciaux and Séverine Adans. Some changes have been done about the recognition, allowing the separation of symbols which were not recognized after the treatment of the original picture, or, on the contrary, the junction between some characters which should constitute some specific symbols. Furthermore, the process of recognition is now based on different data bases depending on the font used. The major contribution of this work concerns the reconstruction of the formula, which is, now, very performing with picture containing Latex characters, but which could be improved for other fonts. This reconstruction is based on the level and the size of analysed symbols.

Date of Award	30 Aug 2010
Original language	French
Supervisor	Marcel Remon (Supervisor), Timoteo Carletti (Jury), Andre Hardy (Jury) & Philippe TOINT (Jury)