To deal with illumination variations in face recognition, a novel two-stage illumination normalization method is proposed in this paper. Firstly, a discrete cosine transform (DCT) is used on the original images in logarithm domain. DC coefficient is set based on the average pixel value of all the within-class training samples and some low frequency AC coefficients are set to zero to eliminate illumination variations in large areas. Secondly, local normalization method, which can minimize illumination variations in small areas, is used on the inverse DCT images. This makes the pixel values on the processed images be close to or equal to that of the normal illumination condition. Experimental results, both on Yale B database and Extended Yale B database, show that the proposed method can eliminate effect of illumination variations effectively and improve performance of face recognition methods significantly. The present method does not demand modeling step and can eliminate the effect of illumination variations before face recognition. In this way, it can be used as a preprocessing step for any existing face recognition method.
JavaScript jest wyłączony w Twojej przeglądarce internetowej. Włącz go, a następnie odśwież stronę, aby móc w pełni z niej korzystać.