Face verification based on Adaboost Learning for Histogram of Gabor Phase Patterns (HGPP) selection and samples synthesis with quotient image method

Abstract. Face verification technology is widely used in public safety, e-commerce, access control, and so on. We propose a novel face verification approach, which combines a relatively new object descriptor—Histogram of Gabor Phase Patterns (HGPP), AdaBoost Algorithm selecting HGPP features and learning binary classifier, and Quotient Image method synthesizing face images under new illumination conditions. Although Gabor wavelets have been widely used in face recognition, previous studies mainly focus on the magni-tude information of Gabor feature, while neglect the phase information of it. We use HGPP as an attempt to utilize the neglected Gabor phase information in face verification. Then AdaBoost algorithm trains binary classifiers, meanwhile significantly reduce the dimension of HGPP. Further, the novel strategy that synthesizes and extends training samples with Quotient Image method enhances our algorithm’s robustness for illumination variation. Experiments demonstrate our novel approach is able to achieve promising face verification results under different illumination conditions