In this paper a new robust image watermarking algorithm based on blocks classification and visual cryp- tography (VC) is presented. First the original image is decomposed into non-overlapping blocks. Then, we use canny edge detection and the support vector machine (SVM) classification method to categorize these blocks into smooth and non-smooth classes. The VC technique is used to generate two image shares: A master share that is constructed according to the block classification results and an owner share gener- ated by using the master share together with a binary watermark. To verify the ownership of the image, the watermark can be retrieved by stacking the master share and the owner share. By skipping blocks which are not robust against attacks, the robustness of our proposed watermarking method is signifi- cantly improved. Our method is completely imperceptible because the watermark pattern is concealed without modifying the original host image.
