Image Authentication based on Q-FMT via Blind Geometric Correction

In the proposed method, the quaternion Fourier Mellin transform (Q-FMT) is used in image hashing, due to its inherent properties of robustness against rotation, scaling, and translation. The input RGB color image is converted into a quaternion domain, which provides the spatial as well as chromatic information. The exemplar-based saliency detection approach and blind geometric distortion correction approach have been also used to identify the tampered regions, in the case of geometric distortions and tampering, that occur simultaneously. The experiment carried out on a large database and shows the proposed image hashing is robust to most digital operations and provides good discrimination. The receiver operating characteristic shows the efficacy of the proposed method.