Geometrically invariant watermarking based on nonlinear anisotropic diffusion

A novel geometrically invariant watermarking scheme based on nonlinear anisotropic diffusion (NAD) detector of the feature points in images is proposed. The NAD feature points detector performs feature points detecting more accurately than Harris detector in an image, which correspondingly can definitely enhance the robustness and the imperceptibility of the watermarking scheme. According to the feature points detected by the NAD detector in watermarking carrier image, Delaunay tessellation method is used to segment the watermarking carrier image into a set of arbitrary shape disjoint triangles. After figuring out noise visibility function (NVF), an adaptive watermarking scheme is proposed based on Human Visual System (HVS). By using this scheme, a right-angled isosceles watermarking image, after being mapped, is embedded into the triangles of the carrier image, which is obtained using the Delaunay tessellation method. In this scheme, the embedding intensity is also well controlled by an adaptive parameter which will vary with the different texture of the carrier image.