基于SIFT特征点匹配的水印图像几何校正算法*

为了有效实现水印图像的水印检测与水印嵌入之间的同步,提出一种应用SIFT(scale invariant feature transform)特征点匹配估计水印图像几何变换参数的校正算法.用两个稳定而又相距最远的图像SIFT特征点的变化估计图像所经历的旋转和缩放参数,校正之后,再用一个稳定的SIFT特征点的变化校正图像...

New watermarked image geometric correction algorithm based on SIFT feature points matching

In order to achieve the synchronization between Watermark detection and watermark embedding, A method that estimate watermarked image geometric distortion factors and corrects it by using SIFT feature points matching is proposed. Test image rotation and scaling parameters could be estimated by the difference between the stable and the farther SIFT feature point's coordinates of it and that of the original image. After revising, the test image translation could be corrected by the change of one the stable SIFT feature point's coordinates of it and that of the original image. Theoretical analysis and experimental results show that the algorithm has high correctional precision, stable performance, and it can recover watermark synchronization of watermarked image by rotation, scaling and translation attacks effectively, so that the watermark could be correctly detected and extracted.