[an error occurred while processing this directive] [an error occurred while processing this directive]

University of Saskatchewan, Department of Computer Science

2001-2002 Seminar Series

Texture Analysis of Images using Weighted Finite Automata

Mark Eramian
Depaartment of Computer Science
University of Western Ontario
London, Ontario

DEPARTMENT SEMINAR
DATE: Friday, May 10, 2002
TIME: 3:30pm
PLACE: Anthropology 132
*** Everyone is welcome ***

Abstract

It has been shown that image compression algorithms based on weighted finite automata (WFA) achieve excellent image compression rates. In these algorithms, greyscale images are encoded as WFA from which the original images can be reconstructed by running finite-length words on the automata. A mapping from words to image pixel coordinates allows the weight with which is a word is accepted by a WFA to be translated into a corresponding greyscale pixel value.

In this talk, we are not so much interested in the compression that WFA can achieve, but instead wish to take advantage of the multiresolution information contained within the WFA representation of an image -- a feature not present in most texture segmentation and analysis methods that have been proposed.

After introducing the basics of image to WFA conversion (and vice versa) we present a method by which local image texture information can be inferred on-the-fly while decoding the automaton. We will show experimental results with particular emphasis on applications to remote sensing, and conclude with some open questions and directions for future research.

About the speaker

Mark completed a BSc in Honours Computer Science and Geophysics, with distinction, from the University of Western Ontario in 1997 at which time he received the Governor General's Silver Medal. He began as a Masters student at Western in 1997 and was approved for transfer into the PhD program a year later. He will have completed his PhD in Computer Science from Western in September 2002. [an error occurred while processing this directive]