Holistic methods for face recognition require lengthy computations because the face image is treated as a vector in a high-dimensional space. In contrast, analytic methods compute a set of geometrical features from the face, then match those features to a database of candidate faces. The authors describe a hybrid method that uses an analytic method to locate boxes containing the eyes and eyebrows, the nose, and the mouth. These three windows are then used as the basis for a correlation match to faces selected from the face database. The selection of faces from the database uses as the search criteria 15 points identifying salient features in the image and their relative spacing. Additional interesting aspects of the method are its ability to rotate images that are not presented “face forward” and the use of an experimentally determined parameter that allows for the fact that faces are somewhere between flat and cylindrical.