Eyeglasses verification by support vector machine

Abstract

In this paper we propose a method to verify the existence of eyeglasses in the frontal face images by support vector machine. The difficulty of such task comes from the unpredictable illumination and the complex composition of facial appearance and eyeglasses. The lighting uncertainty is eliminated by feature selection, where the orientation and anisotropic measure is chosen as the feature space. Due to the nonlinear composition of glasses to face and the small quantity of examples, support vector machine(SVM) is utilized to give a nonlinear decision surface. By carefully choosing kernel functions, an optimal classifier is achieved from training. The experiments illustrate that our model performs well in eyeglasses verification.