Speech and Vision Lab

A representation based on the phase of analytic image is proposed to address the issue of illumination variation in face recognition task. The problem of unwrapping in the computation of analytic phase is avoided by using trigonometric functions of phase. Template matching is used to compare the functions of analytic phase for face recognition. For template matching, the functions of the analytic phase are compressed using eigenanalysis. Performance of the face recognition is improved by using weights derived from the eigenvalues in the template matching.