基于SURF特征区域的鲁棒水印算法

A Robust Watermarking Algorithm Using SURF Feature Regions

针对遥感影像的安全保护,提出了一种基于SURF特征区域的鲁棒水印算法。首先在影像中检测SURF特征点,选取鲁棒性强的点构造互不重叠的特征区域并进行归一化处理,进而对归一化的区域实施整数小波变换,根据生成的模板选取相应位置的系数,最后通过量化的方式将水印信息重复嵌入到每个特征区域内。水印的检测可直接在攻击后的影像中进行,无需校正恢复,实现了盲提取。采用边缘像素替换的策略来保证影像的数据精度,不影响遥感影像的使用。实验表明,该算法具有良好的不可见性,且对常规攻击和几何攻击均具有较强的鲁棒性,可有效保护遥感影像的安全。

To study the security protection of remote sensing images, a robust watermarking algorithm based on SURF feature regions was proposed. Initially, feature points were extracted from original image by using SURF detector, and points with stronger robustness were further selected to construct non-overlapping feature regions and image normalization was performed. Then integer wavelet trans- form was applied on the normalized regions, and corresponding position coefficients were selected ac- cording to the generated template. Finally, the watermark was embedded repeatedly into every feature region using quantization method. Watermark detection could be done directly from the attacked im- age without correction, and also realize the blind extraction. An edge pixels replacement scheme was used to ensure the data accuracy, so the watermark had no effect on the application of remote sensing images. Experimental results show that the proposed algorithm is not only invisible, but also robust against common image processing operations and geometric attacks, so it can protect remote sensing images security effectively.