Fondazione Bruno Kessler - Technologies of Vision

contains material from
Template Matching Techniques in Computer Vision: Theory and Practice
Roberto Brunelli 2009 John Wiley & Sons, Ltd

Chapter 6
Matching Variable Patterns

While finding a single, well defined shape is useful, finding instances of a class of shapes is even more useful. Intraclass variability poses new problems for template matching and several interesting solutions are available. This chapter focuses on the use of projection operators on a one-dimensional space to solve the task: synthetic discriminant functions (SDFs). These projection operators can be optimized to control off peak filter response balancing the requirements of sharp peaks and limited sensitivity to noise while maintaining a constant response over the patterns of a given class.

keywords: synthetic discriminant function, filter off peak response, non-orthogonal image expansion, least squares synthetic discriminant function.

 6.1 Maximizing SNR over class samples
 6.2 Multi-class Synthetic Discriminant Functions