Comparison of illumination normalization techniques for face recognition ORUE, J. P. M. ; CASTRO JUNIOR, A. A. ; GONÇALVES, W. N. X Workshop de Visão Computacional, 2014. p. 275-280.
Face recognition has become one of the most popu- lar areas in computer vision due to its successful applications. In recent years, major advances have been reported in face recognition, and therefore the methods are becoming increasingly accurate and efficient. Despite recent advances, the performance of face recognition methods is still affected by different lighting conditions. This paper presents a comparison of three illumi- nation normalization techniques for illumination invariant face recognition. Experimental results demonstrated that these tech- niques significantly improve the performance of the EigenFaces method. Particularly, the Difference of Gaussians filtering-based normalization technique provided satisfactory results even for feature vectors with low dimensionality (e.g. n = 10).