| In this dissertation, we present the results of our work that seeks to negotiate the gap between low-level features and high-level concepts in both content-based image retrieval and web document retrieval. This work concerns a technique, Latent Semantic Indexing (LSI), which has been used for textual information retrieval for many years. In this environment, LSI is used to determine clusters of co-occurring keywords, sometimes, called concepts, so that a query which uses a particular keyword can then retrieve documents perhaps not containing this keyword, but containing other keywords from the same cluster. In this dissertation, we first examined the use of this technique for content-based image retrieval, using various visual features, namely, global color histogram, subimage color histogram, and color anglogram to represent the image contents. LSI is used to transform the image feature representation into a semantic space. The transformed representation of the images in this lower-dimensional space captures the underlying semantic structure of image contents better than the original feature representation by finding the correlation of low-level features and high-level concepts. We have also examined the use of the LSI technique for web document retrieval in a similar process, using both keywords and image features to represent the documents. Two different approaches to image feature representation, namely, color histogram and color anglogram, are adopted and evaluated. Experimental results show that LSI, together with both textual and visual features, is able to extract the underlying semantic structure of web documents, thus helping to improve the retrieval performance significantly. Based on these research works we firmly believe that negotiating the semantic gap between low-level features and high-level concepts using latent semantic indexing is a promising and feasible approach to improving content-based retrieval, and thus, developing more effective and more intelligent multimedia content management systems. |
