Research Assistant: M. Annamalai
Sponsor: NASA
Content-Based Retrieval:
Digital libraries and the information highways of the future will greatly enhance the ability of users to access materials and to share knowledge and information. Digital libraries contain documents which are text, images, and sound in a single organized package. They make a vast amount of information available to users and there is a need for information retrieval techniques to extract relevant data. Research organizations like NASA have a large amount of data collected in various fields. This data can be used by students, teachers, and researchers if they have an interface to access and retrieve data.
Existing systems like WAIS concentrate on text retrieval. Digital library documents can be accessed using image retrieval too. Digital libraries, unlike traditional libraries, can support an interactive retrieval and retrieval based on imprecise queries. We are conducting research in techniques for image retrieval and retrieval based on fuzzy, imprecise queries. We are investigating techniques in partial content-based retrieval. One technique is to analyze the images to obtain a description based on the semantics of the image. Another technique is to match pieces of images based on a similarity measure. We will build a prototype system to implement and evaluate different techniques with respect to their feasibility in a digital library.
Efficient Image Transmission:
Retrieving large data objects distributed over geographically separated datasites (of the order of hundreds of thousands of kilobytes) takes several seconds and occasionally minutes. We are studying the possibility of reducing the size of the data object by "losing" some of the data using lossy compression techniques before the transmission. The other approach we are studying is to use fast but unreliable protocol such as UDP. We are conducting experiments to investigate these techniques and evaluate their performance.