Self-Similar watermarks for counterfeiting geometrical attacks

This manuscript introduces the concept of self similar watermarks for the purpose of counterfeiting geometrical attacks on copyright protected still images. Geometrical transformations result in the de-synchronization of watermark detection mechanisms enforcing the adoption of exhaustive search strategies for optimizing the algorithm’s efficiency. The self-similar structure attained by the presented watermarks reduces the search ranges and enhances the detection performance by accelerating the overall procedure and minimizing the false detection probability. Various embedding domains (DFT, DWT) were utilized for hiding the watermark information while the presented schemes do not require the original image during detection. Self-similar watermarks prove rather robust against many kinds of attacks, such as linear and non linear filtering, jpeg compression, scaling, cropping and rotation, especially when combined with a robust transform domain.