Research On Zero Watermarking Techniques For Images In Transform Domains

With the rapid development of the Internet and multimedia technology, more and more digital media products become available through different online services. The rapid growth of the multimedia services has created a potential demand for the protection of ownership since digital media can be easily reproduced and manipulated. Digital watermarking technique has been introduced to solve such problems. As to robust image watermarking schemes, all the current schemes need to find a tradeoff between the robustness of the watermarks and the invisibility of the watermarks. The modification of original images is inevitable during the watermarking embedding process although the watermarks embedded are invisible to human eyes. As a result, the copyright protection of images with no data modification is impossible with state of art robust image watermarking schemes.Zero watermarking technique has been presented to solve such problems under the framework of robust watermarking. With no need to embed watermarks into images, zero watermarking technique extracts the stable features of images, and then combines the image features with the copyright watermarks to generate zero watermarks, which are registered in a third-part intellectual property rights (IPR) database for copyright protection.In this paper, zero watermarking techniques for digital images in transform domains have been studied. The transform domains are DFT domain, SVD domain, Contourlet domain and DCT domain respectively. The main contribution of the paper is as follows:1. A zero watermarking scheme is proposed based on the phase features of the DFT coefficients of images. The original images can be reconstructed with only the phase coefficients of DFT of images. When images are attacked in watermarking systems, the low frequency spectrum of DFT is stable while the middle frequency and high frequency spectrum of DFT are unstable. Firstly, the proposed scheme extracts the image feature from low frequency phase spectrum of DFT; secondly, Logistic map is used to generate a random phase sequence; at last the scheme generates the zero watermark bits from the inner products of the two phase sequences. In order to resist global geometric transformation attacks, the scheme makes use of geometric correction. Based on generalized Radon transformation, an improved method is presented to compute the rotation and scaling parameters. The experimental results show the effectiveness of the proposed scheme.2. A zero watermarking scheme is proposed based on the stability of singular vectors of SVD. In current SVD-based zero watermarking schemes, image features are constructed from the biggest singular values of image blocks, which represent mainly the enegy of image blocks. But the singular vectors mainly determine the geometric structures of image blocks, which determine the perceptual characters of images. Research shows that only the first vectors in singular matrices are stable under watermarking attacks. Detailed regions are more robust to attacks compared to smooth regions in images, so the proposed scheme selects detailed regions firstly based on the variances of image blocks; secondly, the scheme generates image feature from the singular vectors U1and V1; at last the image feature bits are scrambled according to the ordering of a Logistic map. The experimental results show the effectiveness of the proposed scheme.3. A zero watermarking scheme is proposed in Contourlet transform domain. DWT has advantages over other transforms because of its multi resolution representation and time-frequency localization, but it lacks in capturing directional information which is important in image processing applications. Contourlet transform has the property to capture directional information as well as inheritance of the multiresolution and time-frequency local characteristics of wavelet. After multi level Contourlet transform, the proposed scheme extracts image features by comparing two orthogonal directional subbands in lowest directional subbands. Block selection procedure is adopted to enhance the robustness of the scheme, and Logistic map is used to ensure the security. The experimental results show the effectiveness of the proposed scheme.4. A zero watermarking scheme is proposed in DCT domain. Studies on the distributions of the DCT coefficients for images show that the Laplacian distribution is the dominant choice balancing simplicity of the model and fidelity to the empirical data, and the width of the Laplacian distributions decreases when we move to higher frequency subbands. A block complexity metric is defined based on the Watson model of HVS (human visual system) defined in DCT domain. The proposed scheme selects complicated blocks from images by use of the complexity metric, and then extracts image feature by comparing the low frequency DCT coefficients of the selected blocks. Also the scrambling of feature bits is performed to enhance the security of the proposed scheme by means of Logistic map. The experimental results show the effectiveness of the proposed scheme.