The paper examines different approaches to classification and discrimination applied to two-dimensional grey-scale images of faces. The aim is the localization of the mouth.
Template matching is compared with results of standard classification methods. We discuss the choice of a suitable template and inspect its robustness aspects.