COLLOQUIUM Department of Computer Science and Engineering University of South Carolina Image Segmentation by Boundary Detection: A Graph-Theoretic Approach Song Wang Department of Electrical and Computer Engineering University of Illinois at Urbana-Champaign Date: April 12, 2002 (Friday) Time: 9:00-10:00AM (Note Morning Time) Place: Swearingen 1A03 (Faculty Lounge) Abstract Image segmentation aims at partitioning an image into a set of disjoint regions. Accurate image segmentation is of fundamental importance in computer vision and image processing. In this presentation, I will introduce a new graph-theoretic approach to image segmentation based on Ratio Cut. This approach is able to detect boundaries and regions simultaneously. Compared with prior approaches, this approach introduces no boundary length, region size, or region shape biases. This allows boundaries to be more accurately located. In addition, I show that the global optima of the proposed ratio-cut cost function can be found in polynomial time. Finally, I demonstrate the utility of the approach by segmenting a set of real (medical and natural) images. Song Wang received his B.S. and M.S. degrees from the Department of Automation, Tsinghua University, in Beijing, China, in 1994 and 1998 respectively. Since 1998, he has been a Ph.D. student in the Department of Electrical and Computer Engineering, University of Illinois at Urbana-Champaign (UIUC). He expects to receive the Ph.D. degree in July 2002. Since 1998, he also has been working as a research assistant in the Image Formation and Processing group at the Beckman Institute of UIUC. His research interests include computer vision, signal/image processing, and pattern recognition. His Ph.D. thesis research focuses on image segmentation. He received C. R. Allen International Student Award in 1999 and Beckman Graduate Fellowship in 2000. He is a student member of the IEEE.