Automatic pose correction for local feature-based faceauthentication

Abstract. In this paper, we present an automatic face authentication system. Accurate segmentation of prominent facial features is accomplished by means of an extension of the Active Shape Model (ASM) approach, the so-called Active Shape Model with Invariant Optimal Features (IOF-ASM). Once the face has been segmented, a pose correction step is applied, so that frontal face images are synthesized. For the generation of these virtual images, we make use of a subset of the shape parameters extracted from a training dataset and Thin Plate Splines texture mapping. Afterwards, sets of local features are computed from these virtual images. The performance of the system is demonstrated on configurations I and II of the XM2VTS database