High-level Semantics of Images in Web Documents Using Weighted Tags and Strength Matrix

The multimedia information retrieval from World Wide Web is a challenging issue. Describing multimedia object in general, images in particular with low-level features increases the semantic gap. From WWW, information present in a HTML document as textual keywords can be extracted for capturing semantic information with the view to narrow the semantic gap. The high-level textual information of images can be extracted and associated with the textual keywords, which narrow down the search space and improve the precision of retrieval. In this paper, a strength matrix is being proposed, which is based on the frequency of occurrence of keywords and the textual information pertaining to image URLs. The strength of these textual keywords are estimated and used for associating these keywords with the images present in the documents. The high-level semantics of the image is described in the HTML documents in the form of image name, ALT tag, optional description, etc., is used for estimating the strength. In addition, word position and weighting mechanism is also used for further improving the association textual keywords with the image related text. The effectiveness of information retrieval of the proposed technique is found to be comparatively better than many of the recently proposed retrieval techniques. The experimental results of the proposed method endorse the fact that image retrieval using image information and textual keywords is better than those of the text based and the content-based approaches.