Construction of a recognition algorithm based on the assessment of the interdependence between local elements of the face image

This study proposes a family (model) of recognition algorithms based on the assessment of the interdependence between local elements of a face image. The main idea of the proposed algorithms is to calculate two-level membership estimates for the considered face image. At the first level, the degree of belonging of the facial element under consideration (for example, the nose) to the given classes is determined. At the second level, a decision is made whether the considered face image belongs to one of the given classes based on the generalization (integration) of estimates of the belonging of the classified facial elements to the given classes. The proposed algorithms are implemented on the ORL and FERET databases. The implementation of these algorithms improves the accuracy and speed of face recognition.