Template Protection for Multimodal Biometric System

Facial image authentication is an important research topic in media security for its wide use in many fields. In this paper, a facial image authentication and tamper detection scheme based on a novel perceptual feature expression named Gabor skeleton is proposed. The Gabor skeleton is constructed from the direction of the maximum Gabor correlation and its absolute magnitude of each block of facial image which expresses the structure feature of the facial image from the point of view of the human vision perception. Then a robust image hash sequence is derived from the Gabor skeleton whose security is ensured by involving a secure key. This hash sequence possesses the capability of supervising the difference of the perceptual content of facial and locating the tamper position. The experimental results demonstrate the feasibility of proposed scheme.