Author : Fatemeh Adim, Sajjad Tavassoli [ Special \\\"9thSASTech-2015, Iran\\\"-2015] [Page No : 1-6] [2015]

In the past decade, digital watermarking have been considered as an important tool for protecting digital data. Due to the high value of images, in recent years, particular attention has paid to image watermarking. A novel image watermarking method in multiwavelet domain based on least square support vector machines (LS-SVMs) is proposed in this paper. In proposed method, according to the special structure and property of image in multiwavelet domain, watermark bits are embedded into multiwavelet approximation sub-bands of image blocks, and a mean value modulation method is employed to modulate their multiwavelet coefficients. In order to robustly extract watermark, LS-SVMs is used to learn mean value relationship between these approximation sub-bands. Due to powerful learning ability and good generalization ability of LS-SVMs, watermark can be correctly recovered under several different attacks.