Digital image watermarking based on three levels discrete wavelet transform to achive robustness against geometrical attacks

Abstract

The availability of the Internet and the growth in using its services has simplified and highly affected the transmission, exchange and acquisition of multimedia information. As a disadvantage of this event, the chances of piracy have increased and became easier. In order to overcome this problem, techniques for identification, authentication and protection of multimedia information are required. Digital image watermarking as a method that provides ownership authentication and copyright protection can be used to protect against piracy. One of the important issues that different image watermarking methods are facing is to have the robustness against different categories of image distortions especially those in geometrical or transform category. Also, since achieving robustness is in contrast with capacity and quality, it is more difficult to offer a method that satisfies all the three requirements. So this project is aimed to propose a method that can offer high resistance against some image distortions in the geometrical attack category while maintaining capacity and quality. The method which is proposed uses discrete wavelet transform (DWT) in combination with singular value decomposition (SVD) in order to achieve high robustness against geometrical attacks. The DWT is performed in three levels on both cover and watermark images and mid and high frequency subbands are selected for hiding the watermark data. Also in order to have a reliable method against the false positive problem, the principal components of the watermark are calculated to be embedded in cover image. The proposed method robustness is tested against scaling, rotation and crop attacks. The experimental results demonstrate that the proposed method has a high robustness against scale attack which is almost 99.9%. Also, robustness against crop and rotation attacks has increased by 14 and 50 percent, respectively, compared to two recent methods. In addition, the capacity for watermark embedding is increased compared to recent proposed methods while the quality of watermarked image is maintained.