March 3, 2000
We examine image understanding from a statistical communications point of view. The goal is to interpret returns from a remote sensor in order to determine information about the source being imaged. Images are modeled through the deformable template framework in which a set of transformations of a prototype is allowed. A prior probability distribution on the source is introduced. The channel corresponding to the remote sensor generates the observable images reflecting projection and noise, and a statistical model via the likelihood function describes the channel operation. Estimation of the source image from the channel returns provides a coherent analytical approach to object recognition and image understanding. Examples will be given associated with object recognition in computer vision and anatomical shape representation in computational anatomy.
Dr. Michael I. Miller became the Director of the Whiting School of Engineering, Center for Imaging Science in July 1998. Prior to that he was the Newton Professor of Biomedical and Electrical Engineering at Washington University in St. Louis. Dr. Miller received his Masters in Electrical and Computer Engineering in 1976 and his Ph.D. in Biomedical Engineering in 1983 from the Johns Hopkins University. He was the recipient of the National Science Foundation's Presidential Young Investigator Award in 1985 and the Paul Ehrlich Graduate Award in 1983. He is co-founder of two companies working in medical image analysis and image understanding. Dr. Miller is currently the Director of the National Center of Imaging Science, a consortium of seven universities working in the areas of automated target recognition and image understanding and is a scientific partner in the National Partnership for Computing Infrastructure (NPACI).