Biography: Wm. Douglas Withers is a professor in the Department of Mathematics of the United States Naval Academy. He received a Ph.D. from the Georgia Institute of Technology in 1983. His research interests include image processing, data compression, fractals, and dynamical systems. He also worked as a research scientist with Iterated Systems, Inc. and Accusoft Imaging Corporation. He earned a Bachelor of Science in Mathematics from Texas A&M University in 1978.
Speech Title: Rapid line parametrization with a hybrid Hough-moment approach
Abstract: The Hough transform is perhaps the best-known method for locating and parametrizing one or more straight lines in an image. Line parametrization is a preliminary step in automated identification of objects in images such as bar codes, cloud-chamber bubble trails, or more complex objects. Although robust with respect to image noise and competing structures, the Hough transform is computationally expensive from the standpoint of time and space. The moment-based method of line parametrization presents complementary strengths and weaknesses to the Hough transform, rapidly computable with almost no memory requirement, but yielding a single line result. This talk will present a hybrid Hough-moment approach to parametrizing lines in an image which combines the strengths of both approaches. The method locates one to several lines in an image with high precision using a small fraction of the resources needed by the Hough transform. We present as a sample application locating and rectifying perspective-distorted documents within an image.