Digital Image Watermarking Resisting to Geometrical Attacks
抵抗几何攻击的数字图像水印

In this paper a watermarking scheme for digital image that can resist geometrical attacks is proposed. The watermark is a key-dependent random vector according to Gaussian distribution. After preprocessed and periodized, the watermark is embedded into cover image in spatial domain exploiting the properties of HVS. The cover image is not needed in watermark detection process. First, compute the Autocorrelation Function(ACF) of estimated watermark. The peaks of ACF are used to determine the geometrical attacks applied to the stegoimage. Experimental results show the scheme is robust to compression, filtering, cropping as well as geometrical attacks.