Human Face Identification from Profile Projection

Abstract

This paper presents face recognition from front face. The idea is to extract the facial profile from a frontal view to provide a measurement for automatic face recognition. The profile is derived from the intensity of horizontal projection of the face image and using profile measure such χ 2 distance for face identification. Three profile projections from 4 images or more are taken for each person, average values are calculated and stored in a database file. Profile measurements are compared with other profile in the database file. Results of experiments to the faces with or without spectacles rotate head (lift, right) and different face expression show that the proposed approach is robust and efficient way of face matching.