Reclassement d'images par le contenu

Georges Quénot (1) , Franck Thollard (1)

1 MRIM - Modélisation et Recherche d’Information Multimédia [Grenoble]

This article presents a method for re-ranking images retrieved by a web scale and state of the art search engine using key words for entering queries. This method uses the visual content of the images and it is based on the idea that the relevant images should be similar to each other while the non relevant images should be different from each other and from relevant images. This idea has been implemented by ranking the images according to their average distances to their nearest neighbors. The re-ranking performed with this method only does not perform better than the original ranking but, when combined to it, it yields a relative performance gain of about 3\% in term of mean average precision and of about 7\% if the complement to 1 of the average precision is considered. This gain is statistically significant and it is brought to a system which is already very effective for exploiting all the textual information available with the images.

Mots clés: Image retrieval; re-ranking; visual content

Domaines: Recherche d'information [cs.IR]

Georges Quénot, Franck Thollard. Reclassement d'images par le contenu. CORIA 2012 - COnférence en Recherche d'Information et Applications, Mar 2012, Bordeaux, France. pp.203-214. ⟨hal-00770631⟩