Web Image Search Based on Ontology-Oriented Hybrid Model

On the Internet, search engines can retrieve images with optimum precision and speed by searching through text and web pages. In the image retrieval systems, Text Based Image Retrieval (TBIR) approach (keywords), Content Based Image Retrieval (CBIR) and the combination of those approaches have been applied. It is worth noting the application of keyboards are user-friendly and hence many researchers prefer to use them in retrieval problems. As expectation, suitable keywords which are much more related to image titles are very important in retrieving images. In this paper, based on WordNet, a model is proposed to retrieve text-based image recognition accuracy and then the proposed approach has been used for clustering (i.e. K-Means algorithm). The semantic web is related to relations and the discovery of meanings and attempts to discover the patterns of the class and the meanings of words. WordNet has a variety of semantic relationships with concepts, therefore it is possible to retrieve file information in the best way. We applied the proposed model on the UIUC data. The results indicated the high performance of the proposed model in evaluating the accuracy of detection of TBIR clustering (> ۹۰% overall accuracy).

Keywords:

Text Based Image Retrieval , Semantic Web , WordNet , K-Means Algorithm