Chapter 8. Recognition of Image Patterns

Summary

The recognition of segmented objects in an image is an extremely important step in image processing. The pattern classification strategies can be grouped as decision theoretic and syntactic classification techniques. Decision theoretic classifiers may again be supervised or unsupervised in nature. Glimpses of all these techniques may be found in this chapter. Both supervised and unsupervised classification techniques have been presented here. Some parametric and non parametric classifiers have been reviewed in a brief yet comprehensive presentation. The concepts and principles of structural pattern classification have been introduced. Finally, the chapter contains discussions on neural network based classifier. Multilayered perceptron with error back propagation technique, which is essentially a supervised learning strategy have been applied in many problems of object recognition. On the other hand Kohonen's self organizing feature map classifies the pattern as an unsupervised classifier. Both these networks have been presented with adequate detail.