Creation of a Large-Scale Image Ontology


This MUSCLE fellowship research project will explore different methods for reducing the "semantic gap" between images and their semantic descriptions and will investigate the way to build a large-scale ontology of images for wide categories of objects. The structure of this ontology, based on lexical resources and generated in the context of OWL (Web Ontology Language) systems (e.g., Protege-OWL Ontolingua, Chimaera, WordNet e.a.), will be used, further, to hierarchically structure a visual catalog that will contain a large number of images of objects. This approach will allow image processing researchers and content-based image retrieval (CBIR) systems to derive signatures for wide classes of objects. These results could also be useful for encoding knowledge on Web pages in order to make it understandable to electronic agents searching for information and could be a step forward for WWW Consortium (W3C) in its objective of developing the Semantic Web.



February 2006 - October 2006


MUSCLE Logo European Commission FP6 Network of Excellence, MUSCLE - Multimedia Understanding Through Semantics, Computation and Learning