Exploring polarmetric infrared using classic image computing methods

In this paper we apply classic image computing methods to a dataset of passive polarmetric long wave infrared data (LWIR). By employing several different pattern recognition techniques such as k-means clustering, support vector machine (SVM), Naive Bayes, and AdaBoost, we demonstrate accurate classification of skin vs. background with an 84% average classification rate. In addition we explored classification using different polarmetric features such as degree of linear polarization (DoLP), angle of linear polarization (AoLP), and the newly introduced, delta and rho functions.