A framework for content-based image retrieval fully exploiting the semantics of annotation

Abstract

We present a framework and an application for semantic-based retrieval of images. Our approach adopts a two-level ontology structure in a subset of OWL-DL. In the core ontology only generic spatial relations are represented, while domain ontologies are specific for the image collection. The approach allows semantic-based relevence ranking and results explanation for query refinement, by exploiting standard and non-standard inferences in Description Logics.