Detection and localization of image tamper with scalable granularity

Image authentication techniques have recently gained great attention due to its importance for a large number of applications. To protect the authenticity of images, some approaches have been proposed in the past years. These approaches include cryptography, semi-fragile watermarking and digital signatures. In this paper, we propose an authentication technique based on what we call characteristic data digest. The characteristic data is derived from the relative difference between the DC value in a central block and its counterpart estimated by adjacent blocks. The size of the block is scalable, from 1x1 to 8x8, depending on the requirements on granularity of tamper localization. Experimental results show that the technique is significantly superior to semi-fragile watermarking based authentication techniques under the same image fidelity, and feature-based authentication techniques, in tolerance range of legitimate processing, and granularity of detection and localization on the tampered area. Because of the simplicity of the algorithm, our technique can be used in real-time image or video frame authentication.