Abstract

A technique is presented for frontal face detection in color images based on facial feature extraction and appearance-based classification. Salient facial features are used to define a search space that is then used in a classification step in order to find the best position of the face in the image. Mouth feature points are identified using the redness property of image pixels whilst eye feature points are detected using a search strategy applied to a subset of regions in a fine region-based segmentation of the candidate face. Face class modeling based on a multivariate normal distribution and discriminating feature analysis is used as the face classification method. The utilization of facial features in this system avoids analyzing the image at every pixel location as well as at multiple scales when detecting faces of different sizes.