Corrupted face image authentication based on noise model

Abstract

In this paper, we propose a method for authenticating corrupted face images based on noise model. The proposed method first generates corrupted images by controlling noise parameters in the training phase. The corrupted image and noise parameters are represented by a linear combination of prototypes of the corrupted images and the noise parameters. With the corrupted image, we can estimate noise parameters of the corrupted image in the testing phase. Then, we can make a synthesized face image from the original face image with the estimated noise parameters and verify it with the corrupted face image. Our experimental results show that the proposed method can estimate noise parameters accurately and improve the performance of face authentication.