基于SIFT特征点的抗几何攻击水印算法

An image watermarking scheme robust to geometric distortion based on SIFT feature

提出了一种基于尺度不变特征变换(SIFT)和伪Zernike矩的抗几何攻击的鲁棒水印算法。首先,利用SIFT算法从载体图像中提取稳定的特征点;然后,根据特征尺度和方向自适应地确定每个局部特征区域大小和方向;最后,从中选择具有较大特征尺度互不重叠的特征区域,并利用量化调制伪Zernike矩幅值方法将水印嵌入到每个局部特征区域内。仿真实验结果表明,该算法不仅具有良好的透明性,而且具有较强的抵抗常规信号处理和几何攻击的能力。

A geometrically robust image watermarking scheme based on scale invariant feature transform （SIFT） and pseudo-Zernike moment is proposed. Firstly, the steady feature points are extracted from the host image by using SIFT; then the size and orientation of each local feature region are ascertained adap tively according to its feature scale and orientation; Lastly, some unoverlapped local feature regions with large-scale are selected for embedding watermarking, and the same digital watermarking is embedded into each local feature region by quantizing the magnitudes of the pseudo-Zernike moments. Simulation results show that the proposed watermaking scheme is invisible, and robust against common signal processing and geometric attacks.