The image of the human face is complex and hard to characterize. This work develops the vocabulary of ridges and parabolic curves, of illumination eigenfaces and elastic warpings for describing the salient features of the face and examines the computer vision problem of face recognition.