Discrete Orthogonal Transforms in Image Analysis - Current Research and Some New Results

R Mukundan

Dept of CSSE, University of Canterbury

Fri Jun 18 15:10:00 NZST 2004 in Room 031, MSCS

Abstract

Image transforms based on orthogonal kernels are commonly found in computer vision applications. In pattern recognition, orthogonal functions are used as shape descriptors of an image, providing minimal information redundancy in a feature set. Image compression algorithms based on transform coding employ orthogonal functions to obtain efficient energy compactness in a set of transform coefficients.

This talk gives a general overview of current research trends in the above areas and introduces some of the discrete orthogonal functions useful in image analysis applications. In particular, the important attributes of discrete Tchebichef polynomials are presented. The results of experiments in invariant pattern recognition, image reconstruction and compression using these functions are also discussed