Improving Eigenface Face Recognition by using Image Registration Preprocessing Methods

This paper investigates using image registration through spatial transformation to incrementally improve eigenface detection in training sets containing non-centered head orientation. In identifying a new face, a training set of faces is created from a scene. After face detection in an image, these faces may exhibit different head orientations. The use of several images combined through image registration offers the possibility of improving eigenface recognition through center normalizing the head orientation of the training sets collected in a dynamic environment.