Since most of present watermarking algorithms for remote sensing images are global watermarking algorithms, it is weak of them to resist rotation, row-column shear, horizontal mirroring and combined geometric attacks. The paper proposes a localized watermarking scheme for remote sensing images based on Harris-Laplace feature region. It generates a random sequence from a key, makes it dual-valued, and selects one paragraph as watermark information to be embedded; then it extracts Harris-Laplace feature points from a host remote sensing image and determines feature regions for watermark embedding according to those points; finally it divides the selected region into rings in accordance with the watermark information and embeds the watermark inside with odd-even quantization method. Experiments have shown that the algorithm show robustness against conventional signal processing such as JPEG compression, noise attacks, filtering, etc. Besides it better resists geometric attacks such as rotation, scaling, moving, or combinations of the above.%针对现有遥感影像数字水印算法多为全局水印算法,抵抗旋转、行列剪切、水平镜像以及联合几何攻击能力较差,提出一种基于Harris-Laplace特征区域的局部化遥感影像水印算法.算法依据密钥生成一个随机序列,对其实施二值化,并选取一段作为待嵌入的数字水印；在宿主遥感影像中提取Harris-Laplace特征点,并依据其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圆环划分,利用奇偶量化方法将水印嵌入其中.实验证明该算法对JEPG压缩、噪声及滤波等常规信号处理具有较好鲁棒性的同时,能够更好地抵抗旋转、缩放、平移及其联合等几何攻击.
展开▼
