Selecting texture discriminative descriptors of capsule endpscopy images

Abstract

In supervised data classification one of the problems is to reduce dimensionality of feature vectors. It is important to find such features which have high ability for discrimination of diverse classes and to get rid of features which are useless for such discrimination. In this paper we propose a new method for feature subset selection utilizing a convex hull (or convex polytope). The method searches for feature space subspaces in which vectors of one class are surrounded by vectors of the other class. The method is applied for selection of color and texture descriptors of capsule endoscope images. The study aims at finding a small set of descriptors for detection of pathological changes in the gastrointestinal tract. The results are compared with results produced by a Radial Basis Function Network classifier.