Presented at the 1998 APIII Conference Return to 1998 Abstract Index
A GRAPHICAL USER INTERFACE FOR CONTENT-BASED IMAGE RETRIEVAL
University
of Pittsburgh
Pittsburgh, Pennsylvania
Lei Zheng, MS
High performance retrieval mechanisms are critical to the effective use of any online database. This is particularly challenging in the case of pathology and other image databases where the sheer volume of terabyte level datasets combined with the need for "query by example" access modes render traditional database indexing methods inadequate. In a collaboration between the Pittsburgh Supercomputing Center (PSC) and the University of Pittsburgh Medical Center (UPMC) we have begun to develop and implement a prototype pathology image atlas which will benefit the medical community by providing powerful tools for pathology image retrieval and diagnosis in both clinical and educational settings. Content-based image retrieval (CBIR) is and essential part of this effort.
This poster session demonstrates the Java-based, cross-platform graphical user interface allowing users to connect to the image database via the Internet. Functions such as query image selection, submission, and prioritized display of retrieved images and associated text descriptions have been implemented. The simple and intuitive user interface is in continued development to support the needs of both users and developers. Because the compute intensive portions of the retrieval process operate on the supercomputer based server at PSC, the client can operate on small PC and workstation class computers.
