We have considered the problem of tracking and recognition using a three dimensional
representation of human faces. First we present a review of the research in the tracking and recognition fields including a list of several commercially available face tracking and
recognition systems. Next, two algorithms are described: one for tracking faces from observed
images and one for recognition of faces from observed geometries. The tracking algorithm uses 3D shape and texture of a human face to estimate the changing position and orientation of a real face in a video image sequence. The recognition algorithm uses principal component analysis (PCA) of range images generated from the 3D shape of a human face to create a database of low-dimensional face representations for ecient recognition. Range images are robust to illumination and texture variations and thus avoid some of the current limitations in face recognition.
