A Hybrid Model of Image Retrieval Based on Ontology Technology and Probabilistic Ranking

There are hundreds of millions of images available on the current World Wide Web. The demand for image retrieval online is growing dramatically. For multimedia documents, the typical keyword-based retrieval method has encountered problems mainly in the areas of: 1) the quality of the search result; 2) the usage of the system. With the advent and development of the semantic Web, information retrieval can widely take advantage of this technology which is expected as the next generation of Internet. However, before shifting up to the semantic Web generation, there are still numerous resources on the current Web without semantic annotation. In this paper, we propose a hybrid retrieval method which is based on the current Web, keyword-based annotation structure, and combining ontology-guided reasoning and probabilistic ranking. A Web application for image retrieval using our proposed approach has been implemented. Furthermore, the system offers recommendations to the user to demonstrate the effectiveness of the model. Experimental results show that the image retrieval recall and precision rates increase by using the proposed hybrid approach