Image Authentication Based on Fractal Features

In this paper, the fractal features of natural images are used to construct an image authentication scheme, which can detect whether an image is maliciously tampered (cutting, wiping, modification, etc.) or not and can even locate the tampered regions. For the original image, the fractal transformation is applied to each of the image blocks, and some of the transformation parameters are quantized and used as the authentication code. The authentication code can be stored or transmitted secretly. To authenticate an image, the new authentication code is computed from the image with the similar method, and then compared with the stored or received code. A metric is proposed to decide whether an image block is tampered or not. Comparative experiments show that the authentication scheme can detect malicious tampering, is robust against such common signal processing as JPEG compression, fractal coding, adding noise or filtering, and thus, obtains competent performances compared with existing image authentication schemes.