An Illumination Pre-processing Method Using the Enhanced Energy of Discrete Wavelet Transform for Face Recognition

Abstract

ABSTRACTAutomatic face recognition is useful in a wide range of applications. The accuracy of a face recognition system is adversely affected due to illumination variations. This study presents an illumination pre-processing method, “Enhanced Energy Discrete Wavelet Transform”, to increase recognition accuracy of front view faces under varying illuminations. This method is implemented as follows. A two-dimensional Discrete Wavelet Transform (2D DWT) decomposes the original image into four frequency sub-bands, namely low-low, low-high, high-low and high-high. The energy of each sub-band is computed. A weighting scheme is employed to increase the energy. The weighting scheme computes weight of sub-band based on its energy and threshold. The new sub-bands are obtained by multiplying sub-band, weight and weighting factor. Then, the new four sub-bands are added to create enhanced energy DWT image. The face recognition is carried out using Gabor magnitude features. The experiments are conducted in facial images of Yale, Pose Illumination Expression (PIE) and Extended Yale B. The results proves that low-low sub-band has maximum weight. The performance analysis is carried out using different threshold and weighting factor values. The weighting factor increases energy of low-low sub-band. The analysis of energy enhancement shows increase in energy of 20.96% and 31.88% in CMU PIE and Extended Yale B, respectively. The increase of energy improves the brightness of image. The recognition accuracy in CMU PIE and Extended Yale B is 99.02% and 99.82%, respectively. The comparative analysis with the state-of-the-art methods proves that the proposed method is an effective illumination pre-processing method.