Institute of Information Science, Academia Sinica

Abstract: In this presentation, we examine a number of research challenges relevant to the IIS and CITI at Academia Sinica. Content-based image retrieval (CBIR) techniques have important applications including image search on the Internet, automatic medical diagnosis, computer-aided drug design, genomic research, multimedia information systems, electronic catalogs in e-commerce, and intelligent transportation systems, just to name a few. The effectiveness of today’s CBIR techniques, however, is still quite limited due to the so-called semantic gap, i.e., the visual features (e.g., color layouts, texture, etc.) used to characterize images are not representative of the semantic content. In this presentation, we first look at how CBIR can be effective for domain-specific applications. In particular, we focus on medical images. We then examine the semantic-gap challenge, and discuss one proposal for this problem, namely the Query Decomposition approach. In terms of video applications, we discuss a live video database management system as a general purpose approach to enable ad-hoc event detection and rapid development of video surveillance applications. The second part of this presentation focuses on wireless multimedia communications. We discuss a Dynamic Stream Merging technique for efficient and robust video-on-demand services to mobile users in wireless mesh networks at the edges. This is a new communication paradigm, in which multicast topologies are created incrementally through dynamic merging of server streams at the mesh nodes. This is accomplished without the assistance of the server. To support high mobility in ad hoc networks, we examine a virtual router concept. Since such virtual routers are stationary, communications through virtual routes are robust regardless of the mobility. Finally, we deal with the communication bottleneck in location-based services. We first review two original solutions known as Safe Region and Resident Domain, and then introduce a highly scalable sensor-based approach. Bio: Kien A. Hua is a Professor of Computer Science, and Director of the Data Systems Lab at the University of Central Florida. He had served as the Associate Dean for Research of the College of Engineering and Computer Science at UCF. Prior to joining UCF, he was a lead architect at IBM Mid-Hudson Lab, where he led a team of senior designers to develop a highly parallel computer system. This system is the precursor to the highly successful commercial parallel computer known as SP2. Dr. Hua received his B.S. in Computer Science, and M.S. and Ph.D. in Electrical Engineering, all from the University of Illinois at Urbana-Champaign. His current research interests include multimedia computing, network and wireless communications, databases, medical imaging, sensor computing, and intelligent transportation systems. He has published widely, with several papers recognized as best papers at conferences. He has served as a Conference Chair, an Associate Chair, and a Technical Program Committee Member of numerous international conferences, as well as on the editorial boards of a number of professional journals. Dr. Hua is a Fellow of IEEE.