基于Sift特征区域的遥感图像水印算法

A Watermarking Algorithm for Remote Sensing Image Based on Sift Feature Region

针对现有遥感图像数字水印算法多为全局水印算法,抵抗旋转、缩放、行列剪切及联合几何攻击能力较差的问题,提出一种基于Sift特征区域的局部化遥感图像水印算法;算法依据密钥生成一个伪随机序列,对其实施二值化,并选取64bit一段作为待嵌入的水印;在截选256×256宿主遥感图像中提取Sift特征点,筛选特征点确定水印嵌入特征区域;最后,根据待嵌入的水印信息,将所选特征区域进行扇区划分,利用奇偶量化方法将水印嵌入其中;实验表明算法对JEPG压缩、噪声及滤波等常规信号处理具有较好鲁棒性,并且能够更好地抵抗旋转、缩放、平移及其联合等几何攻击。

Against most of existing watermarking algorithms for remote sensing image are global algorithms, their resistance to rotation, scaling, ranks of shear and combined geometric attack is poor. It proposes a localized watermarking scheme for remote sensing image based on Sift feature region. It generates a random sequence through the key, and selects a 64--hit paragraph of the values as watermark after binari zation; then it extracts Sift feature points in the 256 × 256 host remote sensing image, filters the feature points and determines feature regions for watermark embedding; finally, it divides the selected region into sectors according to the watermark, and embeds the watermark into it by odd-- even quantization method. Experiments have shown that the algorithm has good robustness against conventional signal processing such as JPEG- compression, noise attacking, filtering, etc. , and it better resists to geometric attacks such as rotation, scaling, moving and their combination.