January 25, 2003 15:46 WSPC/Trim Size: 9in x 6in for Proceedings WIAMIS2002 ROTATION, TRANSLATION AND SCALING INVARIANT WATERMARKING USING A GENERALIZED RADON TRANSFORMATION D. SIMITOPOULOS, A. OIKONOMOPOULOS, AND M. G. STRINTZIS Aristotle University of Thessaloniki Electrical and Computer Engineering Dept. Thessaloniki, Greece A watermarking scheme able to resist geometric attacks is presented in this paper. The wa- termark embedding and detection are performed in a domain which is invariant to geometric attacks such as rotation, scaling and translation. The invariant domain is derived by applying a new generalized Radon transform to the image. The ability of the proposed method to withstand geometric attacks is evaluated experimentally. 1. Introduction Digital image watermarking is a technology which has attracted many researchers in the past few years. A lot of watermarking techniques are laying emphasis on robustness against common digital image processing operations, such as compres- sion or filtering. However it is clear that even the smallest geometric distortions can irreparably harm the detection process of a watermarking scheme due to loss of synchronization between the embedded and the correlating watermark. These geometric distortions usually include translation, rotation and scaling (RST) op- erations which can be easily applied on an image. Recently, many watermarking techniques have been proposed which are ori- ented towards robustness against geometric attacks. In some approaches an ad- ditional pattern is embedded in the image along with the watermark in order to be able to revert the geometric attack 1 . There are also watermarking methods 2 , where the embedding and detection processes are taking place in a domain which is invariant to geometric transformations. The watermarking technique presented in this paper is based on the latter ap- proach for resisting geometric attacks. First, a corner detection scheme detects corners in the image content and finds the most robust among them. This corner is used as an origin for a new generalized Radon transform (GRT) 3 followed by a two dimensional Fourier Transform. Through this sequence of transformations 1