Search space reduction for comic characters retrieval
Clément Guérin, Christophe Rigaud, Karell Bertet, Jean-Christophe Burie, Arnaud Revel, Jean-Marc Ogier
Proceedings of the 19th national congress on Artificial Intelligence and Graphics Recognition (RFIA), 2014
Abstract
Comic strips represent an important piece of cultural heritage in numerous countries and the large-scale digitization offers the possibility of making searches directly in the image content. So far, it is essentially the page structures and the textual contents that have been studied, few works concern the graphic content. We suggest to rely on simple elements, already studied in the literature, such as the position of frames and balloons, to localize more complex elements, such as the comic book heroes, with a higher accuracy. The evaluation of our different contributions from the eBDtheque dataset shows a tail detection accuracy of 81.2%, a character localization accuracy up to 85% and a search space reduction of more than 50%.
Citation BibTeX
@inproceedings{Rigaud2014Localisation,
title = {R{'e}duction de l’espace de recherche pour les personnages de bandes dessin{'e}es},
author = {Gu{'e}rin, Cl{'e}ment} and Rigaud, Christophe and Bertet,Karell and Burie,Jean-Christophe and Revel,Arnaud and Ogier,Jean-Marc},
booktitle = {Proceedings of the 19{\`e}me congr{\`e}s national sur la Reconnaissance de Formes et l'Intelligence Artificielle (RFIA)},
year = {2014}
}