A New Blind Image Authentication for Image Tampering Detection and Recovery Based on Block-wise Feature Classification

This paper presents a blind image authentication technique. It reports on several processes involved in image authentication, tamper detection, and tampered image recovery. The authentication codes of each block consist of three features, variance, grayscale, and edge. To enhance the robustness of each block's authentication codes, these codes will make three copies and permute their data stream before embedding. The proposed blind image authentication method can verify the integrity of an image without requiring access to the original image. Moreover, image recovery for tampered blocks is also conducted with the same authentication code's blocks. The experimental results show that the proposed method can successfully achieve the goal of image authentication and maintain the high visual quality on both of authenticated image and its recovery version if any tamper is happened.