Monday, December 13, 2004
A Visual Concept Ontology for Automatic Image Recognition
A Visual Concept Ontology for Automatic Image Recognition
Authors: Nicolas Maillot, Monique Thonnat and Alain Boucher Date: 2004 Format: PDF
"Multimedia content understanding is expected to play an important role in the future of the Semantic Web. For instance, many image retrieval engines are currently under development. Many of the systems limit their recognition mechanism to low-level image descriptors which are far from semantic notions. On the other hand, other types of system only rely on human annotations [Von-Wun et al., 2002]. We propose an intermediate approach to image understanding. Our approach stems from the fact that experts (e.g. biological and medical experts) of a specific domain often use and share a generic visual vocabulary to describe objects of interest. This paper introduces a domain-independent visual concept ontology which is used as a guide for describing the objects of a domain of expertise. This ontology driven description is planned to support automatic recognition based on image processing techniques."
Authors: Nicolas Maillot, Monique Thonnat and Alain Boucher Date: 2004 Format: PDF
"Multimedia content understanding is expected to play an important role in the future of the Semantic Web. For instance, many image retrieval engines are currently under development. Many of the systems limit their recognition mechanism to low-level image descriptors which are far from semantic notions. On the other hand, other types of system only rely on human annotations [Von-Wun et al., 2002]. We propose an intermediate approach to image understanding. Our approach stems from the fact that experts (e.g. biological and medical experts) of a specific domain often use and share a generic visual vocabulary to describe objects of interest. This paper introduces a domain-independent visual concept ontology which is used as a guide for describing the objects of a domain of expertise. This ontology driven description is planned to support automatic recognition based on image processing techniques."
0 Comments:
