为了更加有效地解决数字水印不可见性的问题, 提出了一种基于视觉显著图的水印方法。对Itti视觉计算注意模进行改进, 提取图像视觉显著图; 利用二级哈尔小波变换方法, 根据图像视觉显著图, 将水印信息嵌入到图像的低频域内; 采用多组SIFT特征点匹配进行精确几何校正处理, 提取水印信息。实验结果表明, 含水印图像的PSNR达到80 dB以上, 提取水印NC值都在0.8以上。因此, 数字水印方法具有极好的不可感知性, 可以有效抵抗噪声、滤波和JPEG 压缩等常规图像处理攻击, 尤其是对任意角度的旋转攻击、缩放、平移等几何攻击具有极强的鲁棒性。

To solve the vulnerable of the digital watermarking, a digital watermarking algorithm is proposed based on the human visual region of interest. Based on the improved Itti visual attention model, image saliency maps are extracted; By using 2D Haar Wavelet Transform, the watermark information is embedded into the low frequency domain image; In the watermark extraction, multiple sets of SIFT feature points are used for precise geometric correction. The experimental results show that the watermark extraction PSNR reach more than 80dB, NC values are more than 0.8. Therefore, the method of digital watermarking has better watermark invisibility, especially for any angle of rotation, zoom, translation and noise; filtering and JPEG compression attacks have strong robustness.