Using Lighting Normalization and SVM for Face Recognition with Uneven Illumination

摘要

In this paper, an efficient face recognition scheme is developed to recognize face in color images with uneven illumination. The proposed scheme comprises two phases, namely face detection and face recognition. For the face detection phase, a lighting normalization function and an isosceles triangle approach are utilized to detect facial regions accurately. For the recognition phase, a SVM scheme is adopted to uniquely identify facial characteristics. The primary advantage of the proposed face recognition system is its ability to handle different facial image sizes under uneven illumination. Moreover, the system performs better than PCA algorithms in term of success rate.