Geometric Invariant Image Authentication System using Hashing
Ram Kumar Karsh · 2020
In the case of image authentication, it has always been a challenging issue to achieve a tradeoff amongst robustness, discrimination, and tampering localization. In this work, this tradeoff has been achieved to a certain extent using the proposed hashing using features such as global and local. The global features are extracted from a modified ring partition approach. Markov absorption probabilities extract the saliency map from the image, which is then used to construct local features. The additional feature, the central orientation information, the distance between image center and furthest non-zero pixel, and image centroid are used to make the proposed system invariant to geometric distortion. The experimental analysis proves the robustness of our approach against content-preserving operations, discrimination, and detection of malicious tampering. The receiver operating characteristics (ROC) depicts the efficacy of the proposed method compared to existing ones.