Research On Robust Digital Image Watermarking Algorithms Against Geometric Attacks

Watermark technology has become a kind of effective means to multimedia information protection and content copyright authentication; it has drawn a lot of attention. One of the key problems of watermarking is the robustness to geometric distortion, which can disable watermark by destroying the synchronization between watermark and image .In this paper, algorithms have been proposed to achieve the robustness of image watermarking against geometric distortion and common signal processing.A robust digital image blind watermarking scheme that combines image feature extraction and pseudo-Zernike moments is proposed. The goal is to resist both signal processing attacks and geometric distortion attacks, such as crop and rotation. We adopt a feature extraction method called Mexican Hat wavelet scale interaction. The extracted feature points can survive a variety of attacks and be used as reference points for both watermark embedding and detection. Several copies of watermark sequence are embedded in the disks which are centered at the extracted feature points. Simulation results show that the watermark can resist geometric attacks like crop and rotation. The watermark is also robust to common frequency-domain attacks, such as JPEG compression, median filtering and noising.A salient feature-based blind watermarking scheme is proposed. It exploits the SIFT techniques to find the robust image feature points , by means of which the watermark can be bound with the image content to improve reliability of the blind detection. In the scheme, we first filter the adaptive points to get the robust feature points, which are used as a reference for both watermark embedding and detection. We embed and detect watermark in a scale and rotation invariable domain. We adopt a mapping method when we compute the domain to reduce the high computational complexity required in rotation and resize computation. Experimental result shows that the proposed scheme is effective when it is confronted with geometric attacks and common signal processing.