Optimal supports for image matching

Abstract

The information theoretic approach provides a foundation for determining new insights and solutions toward image modeling and analysis problems. The underlying principle is that a search through an image can be viewed as a reduction of the expected uncertainty in the classification of the image. Specifically, we propose using the Kullback (1959) relative information for the determination of the support which maximizes the feature class separation, which consequently should minimize the probability of misclassifications. The methods are applied to face detection and two view image matching using internationally available databases.