Fourier-Mellin transform for Robust Image Hashing

Abstract

This paper proposes a new perceptual image hashing technique based on the Fourier-Mellin Transform (FMT). The idea consists of exploiting the invariance property of the FMT to extract robust hashes using overlapping blocks in the transform domain. To secure the system, two secrete keys are used in different stages. The results show good robustness against signal processing operations and geometric attacks including translation and rotation and demonstrate the superiority of the proposed technique over related transform based hashing techniques.