一种强稳健性的抗几何攻击图像水印算法

Geometrical Attack Resistant Algorithm of Digital Image Watermarking with Strong Robustness

数字图像水印的抗几何攻击能力一直是研究的热点和难点。提出了一种基于奇异值分解的抗几何攻击图像水印算法,算法利用了奇异值良好的稳定特性,通过对宿主载体和水印图像的奇异值进行混合运算嵌入水印,同时引入了FastICA算法,在满足算法强稳健性和透明性的同时,保证了快速的收敛性和盲提取要求,属于新观点创新。实验结果表明,算法能很好地抵抗高斯低通滤波、维纳滤波、中值滤波、椒盐加噪、高斯加噪、JPEG有损压缩等多种攻击,对旋转、剪切、缩放、平移、错切几何攻击具有优于传统小波算法无法比拟的强鲁棒性,提取水印与原水印相关系数NC(Normalized Cross-Correlation)平均优于小波算法43.35%,显示了算法的独特优势。

The resist to geometrical attack of digital images watermarking is always the hot and difficult issues. A blind watermarking algorithm was proposed based on the singular values decomposition for its favorable stability. This algorithm embedded a small disturbance watermarking in the image＇s singular values to achieve strong robustness and transparency, and introduced the FastlCA algorithm to ensure fast astringency and blind extraction. The experiment shows that this algorithm has the strong robustness to Gaussian low pass filter, wiener filter, median filter, adding salt ＆ pepper noise, adding Gaussian noise, JPEG compression, etc. The algorithm can also resist the geometrical attack, such as rotation, shearing, shrink ＆ magnify, shifting, cropping. NC （Normalized Cross-Correlation） of the original watermark and extraction watermark is superior to the traditional wavelet algorithm average 43.35% .It shows the unique superiority great advantage.