No abstract

No abstract

Recognizing facial images with nonhomogeneous illumination is a challenging task. Retina and difference of Gaussians filtering have been applied to facial images in order to remove variations in illumination. This paper presents a study on how these preprocessing operations affect or improve the performance of the correlation filters in a face recognition task. These preprocessing operations were applied to CMU and YaleB facial databases, which containing images with homogeneous and nonhomogeneous illumination, respectively. Results show that these operations improve the performance of correlation filters for recognizing facial images with variable illumination.