A new robust semi-blind image watermarking based on block classification and visual cryptography In this paper a novel and robust image watermarking algorithm based on block classification and visualcryptography (VC) is presented. The proposed method inserts a watermark pattern without modifying the original host image. First the original image is decomposed into non-overlapping blocks. Then, we use canny edge detection and support vector machine (SVM) classification method to categorize these blocks into smooth and non-smooth (non-edge and edge) classes. The VC technique is used to generate two image shares. A master share that is constructed according to the block classification results and then owner share is generated by comparing master share together with binary watermark according to the (2,2) VC technique. To verify the ownership of the image, watermark can be retrieved by stacking the master share and the owner share. Experimental results show that the proposed watermarking scheme is completely imperceptible and also has high robustness against common image processing attacks.