STATISTICAL STRUCTURING OF PICTORIAL DATABASES FOR CONTENT-BASED IMAGE RETRIEVAL-SYSTEMS

This letter presents a two-stage statistical approach for ''exploring and explaining'' a pictorial database, for content-based image retriev al systems. First, we describe how correspondence analysis provides im age classes, as well as facilitates the understanding of the role of i mage primitives and attributes used to index pictures. Such understand ing allows an intelligent choice of features, and thus computational s avings, to be made. Second, ascendent hierarchical classification perm its the structuring of the database, in order to ease picture indexing and retrieval.