Extraction of shape feature for image authentication

In this paper, a shape feature using Zernike moments for image authentication is proposed. It can be used as an image hash sequence. At first the color space of the input image is transformed from RGB to YCbCr. Then these three components are mapped to a unit circle by conformal mapping. Then Zernike moments of three image component s are calculated and the amplitudes and phases of modified Zernike moments are connected to form the intermediate hash. Lastly, the final hash sequence is obtained by pseudo-randomly permuting the intermediate hash sequence. Similarity between hashes is measured by a new distance defined in this paper. Experimental results show that this method is robust against most content-preserving attacks. The threshold can be got by robustness and uniqueness tests. The distance of hashes between two different images is bigger than the threshold. And this method can be used to detect image forgery involving structural and color modifications.