ALIGNMENT FREE OCCLUDED FACE RECOGNITION USING SPARSE

Abstract

It is not possible to take picture in same pose and also not a person may have same facial expression. Illumination variation generally causes performance degradation in a face recognition systems under actual environments. Moreover, there may always not possible to full picture of face due to some occlusion some object or by any other person. Therefore, we propose a face recognition system using sparse representation classifier method on efficient facial feature to develop real time face recognition system. In the proposed approach, image is subdivided then LBP and Gabor is applied to face images to find out the features and then sparse representation is applied to recognize the face. The individual matching scores obtained from two sub-images are then integrated using a weighted-summation operation, and the fused-score is utilized to classify the unknown user. Performance evaluation of the proposed system was performed using an extended Yale face database B which consists of 2,414 face images for 38 subjects representing 64 illumination conditions under the frontal pose.