"Étude comparative de trois ensembles de descripteurs de texture pour la segmentation de documents anciens"

Recently, texture-based features have been used for digitized historical document image segmentation. It has been proven that these methods work effectively with no a priori knowledge. Moreover, it has been shown that they are robust when they are applied on degraded documents under different noise levels and kinds. In this paper an approach of evaluating CIFED 2014, pp. 41–56, Nancy, 18-21 mars 2014 42 M. Mehri, M. Mhiri, P. Héroux, P. Gomez-Krämer, M. A. Mahjoub, R. Mullot texture-based feature sets for segmenting historical documents is presented in order to compare several texture-based feature sets. We aim to determine which texture features could be better adequate for segmenting on the one hand graphical regions from textual ones and on the other hand for discriminating text in a variety of situations of different fonts and scales. For this purpose, three well-known and widely used texture-based feature sets (autocorrelation function, Grey Level Co-occurrence Matrix (GLCM), and Gabor filters) are evaluated and compared. An additional insight into the computation time and complexity of each texture-based feature set is given. On more than 300 historical documents, qualitative and numerical experiments are given to demonstrate each texture-based approach performance. MOTS-CLÉS : Documents historiques numérisés, segmentation, texture, multi-résolution.