Geometrically Robust Image Hash with Tamper Detection and Location

Image hashing is an effective approach to image authentication. In this paper, a geometric distortioninvariant image hashing scheme, which can detect and locate the tampering, is proposed. The reported scheme performs Polar Complex Exponential Transform (PCET) on an image, and then the magnitudes and the phases of PCETs are combined to generate a set of rotation and scaling invariant features. To address the issue of security, the PCET moments are weighted by secret key K1 and permuted by secret key K2 respectively. Then the key-dependent transformed feature space is used to calculate the image hash. If the test image is verified as tampered image in the verification process, a map of the estimated tampering can be reconstructed from the image hashes to locate the tampering regions. Experimental results show that this scheme can tolerate almost all the typical image processing manipulations such as geometric distortion, JPEG compression, blurring, noise addition, and enhancement. Besides robustness, the proposed hashing scheme has the ability of detecting minute tampering with localization of the tampered area.