Statistical Engineering Division Seminar

Anuj Srivastava
Department of Statistics
Florida State University
NIST North Room 152
Thursday, February 23, 2006, 10:30 AM

Abstract

Shapes can play an important role in recognizing objects in images. Focusing on shapes of boundaries of objects, we are interested in studying shapes of: (i) closed curves in 2D images, and (ii) surfaces in 3D images. Our approach is to specify a space of appropriate representations, e.g. space of closed curves, and form a quotient space removing all shape preserving transformations. The resulting shape space is an infinite-dimensional, nonlinear space and we use its differential geometry to develop tools for shape analysis. A basic tool is to construct a geodesic path, in the shape space, connecting any two given shapes. Using this tool, we can define shape statistics -- means and covariances, and impose probability models on shape spaces. We demonstrate the use of these two in shape modeling, clustering, and classification. Additionally, we present results on extracting shapes from low-quality images, i.e. segmentation, using a Bayesian framework that utilizes a prior model on shape space.

Speaker Bio

Anuj Srivastava received his B.Tech degree in electronics engineering from the Institute of Technology, Banaras Hindu University, Varanasi, India, in 1990, and M.S. and Ph.D. degrees in electrical engineering from Washington University, St Louis, MO, in the years 1993 and 1996, respectively. He was a visiting research associate in the Division of Applied Mathematics, Brown University, from 1996-97, and then joined Department of Statistics at Florida State University in Tallahassee, FL as an Assistant Professor in 1997. Currently, he is an Associate Professor there. He is also a member of the Center for Applied Vision and Imaging Sciences (CAVIS) at FSU. He was an associate editor of IEEE Signal Processing (2004-06), and currently serves on the editorial board of IEEE Pattern Analysis and Machine Intelligence and Journal of Statistical Planning and Inference. His research interests include statistical computer vision, computational statistics, and the use of differential geometry in statistical inferences. He has published more than 70 journal, conference and book-chapter articles in these areas. He was awarded FSU developing scholar award in 2005.