Skip to Main content Skip to Navigation
Theses

Apport des Graphes dans la Reconnaissance Non-Contrainte de Caractères Manuscrits Anciens

Abstract : Our work has been motivated by the issue of generic handwritten characters recognition. We try to address it with structural methods based on graph modelling. The documents processed are unconstrained and come from different periods. Classical statistical methods are efficients but they can only process languages with restrained vocabulary according to a learning phase. Two recognition systems based on attributed graphs are proposed. The first one uses numerical attributes and random graphs for modelling the learning base. The structural information changes the complexity notion and allows an interesting cooperation with statistical methods. The second one uses hierarchical fuzzy attributes. It is a model-based recognition system with no learning phase. It brings an interesting first step for generic recognition.
Complete list of metadata

Cited literature [174 references]  Display  Hide  Download

https://tel.archives-ouvertes.fr/tel-00267232
Contributor : Denis Arrivault <>
Submitted on : Wednesday, March 26, 2008 - 5:21:35 PM
Last modification on : Friday, October 23, 2020 - 4:36:17 PM
Long-term archiving on: : Thursday, May 20, 2010 - 9:25:05 PM

Identifiers

  • HAL Id : tel-00267232, version 1

Collections

Citation

Denis Arrivault. Apport des Graphes dans la Reconnaissance Non-Contrainte de Caractères Manuscrits Anciens. Traitement du signal et de l'image [eess.SP]. Université de Poitiers, 2006. Français. ⟨tel-00267232⟩

Share

Metrics

Record views

357

Files downloads

843