Affine Invariant Watermark Scheme using Genetic Algorithm

Abstract

Geometrical attacks are among the most challenging problems in present day digital watermark. Such attacks are very simple to implement yet they can defeat most of the existing digital watermark algorithms without causing serious perceptual image distortion. Geometric attacks can very easily confuse the decoder unless it transforms the image back to its original size/orientation, i.e., recover the lost synchronism. To be able do so, the decoder needs to know how the image has been manipulated, i.e., needs to know geometric transformation parameters. In this research, we report a novel method to estimate the geometric manipulation. We compute the point pattern matching measure for the geometric manipulation. The shape-specific points of the original image are computed. The point matching is realized by genetic algorithm. The proposed scheme does not require the original image because SSP information of the original image has been memorized by the neural network. This method has been proved its robustness to geometric attacks through experiments.