Title:
Distance: Measuring Nearness of Collections of Digital Images
Date:
Friday, March 18, 2010 from 2:00pm – 3:00pm
Coffee and light refreshments are served in E2-361 from 1:30 P.M.
Location:
E3-262 EITC Building (Senate Chambers), University of Manitoba Fort Garry Campus
Speaker:
Dr. James Peter
Department of Electrical and Computer Engineering
University of Manitoba
Abstract:
This seminar considers the problem of how to formulate a framework for the study of the nearness of collections of objects such as digital images. The solution to the problem stems from recent work on approach spaces, near sets, and various approaches in determining the distance between points and sets or between sets or between collections of sets.
Each region-of-interest in a digital image can be viewed as a collection of neighbourhoods, where each neighbourhood Nρ(x0, ε) is a set containing a centre x0 and radius ε and all members of a neighbourhood are within epsilon distance of the centre. Distances between elements of a neighbourhood and a centre x0 are determined with a distance function ρ. The collection of all subsets of a set equipped with a distance function satisfying certain conditions is a called an approach space. By viewing digital images as sets of points, we can consider approach spaces as a framework for measuring the nearness of a query image to a collection of images and consider identifying those members in a collection that are descriptively closest to the query image.
The basic approach in this seminar is intuitive with the underlying mathematics shining through the examples that are considered. Since neighbourhoods and approach space theory are important in this work, some attention is given to the basics of these topics. This seminar takes into account the early work on distance by Henri Poincaré, Felix Hausdorff and the more recent work by Robert Lowen and others. In addition, two fundamentally different types of distance (geometric and feature-based distance) are considered along with many examples in image analysis.
Speaker Bio:
Dr. James F. Peters, B.Sc.(Math), M.Sc.(Math), Ph.D., Constructive Specification of Communicating Systems, Postdoctoral Fellow, Syracuse University and researcher in the Rome AI Laboratory, Griffiss Air Force Base, New York (1991), Asst. Prof., University of Arkansas, 1991-1994, and Researcher in the Mission Sequencing and Deep Space-Telecommunications Divisions at the Jet Propulsion Laboratory/ Caltech, Pasadena, California (1991-1994). In 1992, he verified the correctness of the command sequencing rules for the NASA TOPEX/Poseidon ocean-monitoring satellite (it ceased operation in 2007 after 62,000 orbits but still remains 1336 km above the Earth) and, in 1993-1944, he worked on a proof of correctness of an antenna controller for the NASA deep space network. His early work at the U of M included Petri net models of satellite subsystems and the design of robotic inspection systems for Manitoba Hydro transmission lines. He is now a Full Professor in the Department of Electrical and Computer Engineering (ECE) at the University of Manitoba. He is co-Editor (with Prof. Sankar Pal, Indian Statistical Institute, Kolkata, India) of a book on rough-fuzzy image analysis published in 2010 by Chapman & Hall/CRC Press, author of over 40 articles published in refereed journals during the past 5 years, including a best journal article award in 2008 and 2010, best conference paper award in 2007, and 10 journal articles published so far in 2011. He is co-Editor-in-Chief of the TRS journal, Associate Editor of two journals and Editorial Board member of a number of other journals. His main research interests are topology, various forms of sets such as near sets, and image analysis.
Cost:
This will be a free event.
Contact:
For questions or more information: Jun Cai 474-6419
