Robust wavelet-based blind image watermarking against geometrical attacks

Abstract

Given its strong similarity to the human visual system, the wavelet transform has been applied in watermarking successfully. However, due to its sensitivity to rotation, scaling and translation, the wavelet-based watermarking algorithms are vulnerable to geometrical attacks. In this paper, we design a novel wavelet-based watermarking scheme resilient to geometrical attacks using a rotation-invariant log-polar mapping to eliminate all the effects of the geometrical operations in the input image before computing its discrete wavelet transform. A pixel-wise masking method used in the wavelet domain to scale the watermark is applied in order to achieve the best compromise between the requirements of robustness and imperceptibility for the embedded watermark. Experimental results demonstrate the advantages of our scheme, particularly its robustness against geometrical attacks.