Machine Vision, 3rd Edition

CHAPTER 19 Invariants and Their Applications

19.1 Introduction

Pattern recognition is a complex task and, as stated in Chapter 1, involves the twin processes of discrimination and generalization. Generalization is in many ways more important than discrimination—especially in the initial stages of recognition—since there is so much redundant information in a typical image. Thus, we need to find ways of helping to eliminate invalid matches. This is where the study of invariants comes into its own.

An invariant is a property of an object or class of objects that does not change with changes of viewpoint or object pose and that can therefore be used to help distinguish it from other objects. The procedure is to search for objects with a specific ...