Numerical information is volatile and watermarking is a solution to assist copyright protection. During the detection step, the synchronization of the mark is a great problem. Geometric transformations can defeat the detector of the mark by desynchronizing the mark. Our scheme is based on warping of pre-defined triangular patterns. The content of the image (feature points) is used to mark independently different regions. This allows the synchronization of the mark for the detection step. Feature points mixed with a Delaunay tessellation permits to mark each triangle of the image. The detection is performed by warping triangle to a reference pattern and correlating with a reference triangle. Different algorithms have been developed in the spatial domain and in the frequential (DCT) domain. Our results show that our schemes are robust to Stirmark and other geometric transformations on different categories of images.
展开▼