Multimedia retrieval:

While multimedia technology is growing dramatically, the searching of multimedia data objects becomes an important activity. However, this kind of search activity is far more challenging than the searching of the text-based documents. We aim at designing an innovative approach that enables the semantic search of the multimedia objects by the discovery and meaningful indexing of their semantic concepts. Such objects may include images, videos and different forms of music. By analyzing the users' search queries, relevance feedback and selection patterns, semantic concepts can be discovered and migrated through an index hierarchy. Through the growth and evolution of the index hierarchy, the semantic index may be dynamically constructed, validated, and built-up.

Semantic Image Retrieval

With the rapid increase of the volume of digital image collections, image retrieval has been become one of the most research area. The effectiveness of image retrieval depends on meaningful indexing. In our project, we propose an extension of image annotation models which are using knowledge-based expansion and contextual feature-based index expansion with adaptive segmentation of Hue, Saturation and Value (HSV) color space. Our system is evaluated quantitatively using more than 100,000 web im- ages and around 1,000,000 tags. Experimental results indicate that this approach is able to deliver highly superior performance compared with the original approach and simple color based retrieval systems.