基于SIFT抗几何攻击的数字水印算法

Digital watermarking algorithm of anti-geometric attacks based on SIFT

针对数字水印信息易遭几何攻击的问题以及水印算法的不可见性与鲁棒性的平衡性问题,提出一种基于尺度不变特征变换(SIFT)的图像局部特征点的数字水印算法。首先利用SIFT算法在原始图像中寻找局部特征点,再以局部特征点确定多个满足一定条件的圆形局部特征区域(LFA),经LFA正规化后将水印嵌入到LFA的离散余弦变换(DCT)域的中频系数中。其中,考虑到水印信息对图像质量的影响,嵌入强度根据Watson人类视觉模型进行动态调整。实验结果表明,该算法得出的峰值信噪比(PSNR)和水印相似度数值较高,说明该算法不仅保证了较好的水印不可见性,并且在一定的几何攻击下表现出较强的鲁棒性。

To solve the problems that the digital watermark information are vulnerable to geometric attacks and the balance between invisibility and robustness of watermarking algorithm, a digital watermarking algorithm was proposed based on Scale Invariant Feature Transform （SIFF） for image local feature points. Based on SIFT algorithm local feature points, circular Local Feature Area （LFA） that met certain conditions were found. After the LFA regularization, the watermarking was embedded in intermediate frequency coefficient of LFA Discrete Cosine Transform （DCT） domain. Among them, considering the influence of the watermark information on the image quality, the embedding strength was dynamically adjusted according to the Watson human visual model. The experimental results show that the Peak Signal-to-Noise Ratio （PSNR） and the similarity of watermark get higher. The algorithm guarantees better watermark invisibility and robustness performance under certain geometric attacks.