Modèle de langue visuel pour la reconnaissance de scènes

We describe here a method to use a graph language modeling approach for image retrieval and image categorization. Since photographic images are 2D data, we first use image regions (mapped to automatically induced concepts) and then spatial relationships between these regions to build a complete image graph representation. Our method deals with different scenarios, where isolated images or groups of images are used for training or testing. The results obtained on an image categorization problem show (a) that the procedure to automatically induce concepts from an image is effective, and (b) that the use of spatial relationships, in addition to concepts, for representing an image content helps improve the classifier accuracy. This approach extends the language modeling approach to information retrieval to the problem of graph-based image retrieval and categorization, without considering image annotations. MOTS-CLÉS : Représentation de graphes, recherche d’image, catégorisation d’image