基于归一化小波域抗几何攻击的图像水印算法

Image Watermarking Algorithm Based on Geometric Attacks in Normalized Wavelet Domain

在提出的新算法(NBWW)中,利用单位圆内的图像归一化进行几何失真的同步,水印的嵌入和提取在归一化域进行,嵌入时,利用9/7双正交小波对归一化图像进行两级小波分解,通过修改小波系数嵌入水印,同时利用Watson小波域量化噪声模型在不可见性和鲁棒性之间取得折中。实验结果表明,NBWW算法能有效抵抗旋转、缩放攻击,对JPEG压缩、加噪、滤波等常规攻击同样具有很好的鲁棒性和实用性。

In the proposed new algorithm named normalized based image watermarking using wavelet （NBWW）, the geometric distortions are first synehronized by image normalization in unit circle, and the insertion and detection of watermark arc implemented in normalized domain. The 9/7 biorthogonal wavelet is adopted in insertion to do the two-level wavelet decompose of normalized image, and the watermark bits are inserted by modifying wavelet coefficients. Watson＇s wavelet quantization noise model is used to achieve the compromise between invisibility and robustness of watermark. The experimental results show that NBWW algorithm can effeetively resist attacks of rotation and sealing, and has good robustness to common attaeks like JPEG compression, added noise and filtering, and thus suitable for practical applieations.