Geometric attack is a challenging issue in digital image watermarking field .In this paper , we propose a spatial domain-localised robust digital image watermarking algorithm , which takes the image corner as the feature point and is based on odd /even quantified embedding/detection method .The main means the algorithm adopts are: ( 1 ) for robust corner detection , it uses the curvature scale space-based corner detection algorithm with adaptive threshold and dynamically supported area ;(2) it screens corner based on the complexity of image’s texture, and takes the screened corner as the circle centre to form circular feature area with geometric invariance ;(3) it designs in feature area an annular watermarking pattern , and uses in spatial domain the classic odd/even quantification method to embed and detect watermarks .Experiments show that , after the watermarks are embedded , the algorithm can still keep higher quality of image vision , and has quite strong robust on common signal treatment and against the geometric attacks including rotation , zooming, translation, cropping and the composition of them all .%几何攻击是数字图像水印领域中一个富有挑战性的问题。提出一种以图像角点为特征点，基于奇偶量化嵌入/检测的空间域局部化鲁棒数字图像水印算法。采取的主要方法有：（1）采用带有自适应阈值和动态支持区域的基于曲率尺度空间的角点检测算法进行鲁棒角点检测；（2）基于图像纹理复杂度进行角点筛选，并以筛选后的角点为圆心形成具有几何不变性的圆形特征区域，用于水印的重同步；（3）在特征区域设计一种圆环形的水印模式，并在空间域采用经典的奇偶量化方法嵌入并检测水印。实验表明，该算法在嵌入水印后仍能保持较高的图像视觉质量，对常规信号处理和包括旋转、缩放、平移、剪切及其组合在内的几何攻击均具有较强的鲁棒性。
展开▼
