Extracting facial features by an inhibitory mechanism based on gradient distributions

The extraction of facial features is a fundamental and crucial step in most face detection and recognition systems. Here a set of approaches are proposed for extracting internal and external facial features in a gray-level images containing single or multiple faces of various sizes, at different locations without any restriction on the background. These approaches are distinctive in several aspects: the use of some rarely exploited photometric properties as the basis of facial features extraction, a novel metrics on radial symmetry of gradient orientations, and an inhibitory mechanism for extracting internal facial features and a simple mechanism for external ones. Copyright (C) 1996 Pattern Recognition Society.