基于特征点的抗几何攻击水印算法

利用Harris特征点并结合图像的归一化原理提出了一种新的数字水印方案。由于Harris算子的算法简单,稳健性较好,图像的特征点在经过几何攻击后仍然可以保持。而且,归一化的图像对图像的旋转不太敏感,所以首先对每一个以特征点为圆心的互不重叠的圆归一化以确定水印的嵌入点,然后把水印嵌入到原来的图像中,这样,可以很好地解决水印嵌入和检测的同步问题。实验证明,该算法能很好地抵抗如旋转,缩放,剪切等形式的几何攻击与常规信号处理攻击。

Digital image watermarking based on feather point against geometrical attacks

In this paper a digital watermarking scheme is proposed based on a combination of Harris feather point and image normalization.The scheme of feather point detector based on Harris is simple and reliable,the feather points remain holding after geometry attack.The normalized image of an image is nearly invariant with respect to rotations.As a result,the authors apply image normalization to these non-overlapped circles which centered at the feature points,after that,the authors can determine area for watermark embedding.Then the authors embed watermarks in the original circles.In this way,the synchronization issue for watermark embedding and detecting can be best solved.Simulation results show that this method can resist variety of geometry attack,such as rotation,scaling,shearing and other common signal attacks.