一种基于图像特征的稳健水印算法

本文提出一种基于图像特征的抗几何攻击的稳健水印算法。算法首先用Harris检测器提取图像的特征点，进行Delaunay分割的图像分解，然后水印信号经过图像的视觉掩蔽分析嵌入图像的特征区域，水印图像保持了很好的感知质量，算法检测水印不使用原始图像。由于图像的特征点在经过几何攻击后仍可以保持，这样就很好地解决了水印的同步问题，增强了水印的稳健性。实验结果表明算法可有效的抵抗旋转，尺度变换，有损压缩等各种攻击。

Robust Watermarking Algorithm Based on Image Features

The paper presents a digital watermarking algorithm based on image features, which is robust to geometrical distortions. Firstly, the algorithm extracts feature points of the image using Harris detector and performs a Delaunay tessellation on the set of points. Then the watermark signal is embedded in the image feature zone through visual masking analysis. Watermarked image preserves a better perceptual quality. The original image is not needed in watermark detection. Because the image feature points can be preserved after geometrical distortions, the scheme solves the watermark synchronization problem and enhances watermark robustness nicely. The experiment results show that the scheme is robust to different attacks, such as rotating, scaling, lossy compressing, and etc.