Supervised Classification of Thermal High-Resolution Infrared Images

In many classification problems relevant features are unknown a priori. Therefore many candidate features are introduced to represent the phenomenon. Unfortunately it is often true that most of these are either partially or completely redundant to the target. Thus when the size of the dataset is large an important primary step in the classification task is to remove the unwanted features. In this framework this study proposes a new subset selection algorithm called JSS+E (Jackknifed Stepwise Selection with Exhaustive search) in order to improve the stepwise selection procedure. The procedure is applied in a supervised classification approach for the differential diagnosis of Raynaud's Phenomenon on the basis of functional infrared (IR) imaging data. The results discussed for a dataset collected at ITAB laboratory in Chieti allow to refine the experimental protocol in a completely new non-invasive way.