Eigenedginess vs. eigenhill, eigenface and eigenedge

Recognizing human faces, invariant to illumination and facial expressions is a difficult task. This paper proposes a method of face recognition using eigen analysis on edginess-based representation of the face. One dimensional processing of images is used to extract the edginess of the face. Experimental results show that the edginess-based representation performs better than other edge and gray level based representations under variation in illumination and facial expressions. This paper also discusses the effect of enhancing suitable edginess values and elimination of first few eigenvectors on the recognition performance.