Research Of Image Watermarking Algorithms Based On DWT-SVD

As an important branch of information hiding, digital watermarking technique plays animportant role in copyright protection and content authentication of digital multimedia products,and its related theories and algorithms are being constantly developed and improved. Currently,digital watermarking technique has become a hot topic. The problems like contradiction betweenthe embedded watermarking capacity and the robustness and the weak ability to resist geometricattacks have not been perfectly solved till now. Therefore, combining good characteristics of DWTand SVD, several robust watermarking methods are proposed in the thesis. The practicalsignificance and innovation of the research are as follows:1. The thesis brifely introduces the background, application and research status of digitalwatermarking. It summarizes the principles, characteristics, classification of digital watermarking indetail, and emphasizes drawbacks of typical image watermarking methods. In addation, theadvantages of applicating logistic map, DWT and SVD in digital watermarking are introduced.2. In order to solve the contradiction between the embedded watermark capacity and robustness,a robust DWT-SVD watermarking algorithm based on multi-level DCT is proposed. Combining theadvantages of DWT and SVD, and making use of the energy concentration of multi-level DCT,more datas with larger numerical values can be botained to embed the watermark, so the proposednot just can improve the watermark capacity, and can maintain the good robustness.3. Currently, most watermark methods show good resistence to common attacks, but weakability to geometric attacks, so a digital watermarking method based on Zernike moment isproposed. The inscribed circle of the host image matrix is chosen as the Zernike momentscomputation area, and the square of the inscribed circle is selected as the watermarking embeddingarea. The watermark is embedded in wavelet domain by quantifying of the singular value matrix ofeach block; some Zernike moments of watermark image are saved as the key which is used tocorrect the parameter of geometric attacks when the watermark is being extracted. Experimentsresults show that, the proposed not only is robust to any angle rotation attacks and wide range ofscaling attacks by the advantage of Zernike moments, but also has good resistence to commonsignal processing.4. The watermarked image will be severely damaged by the geometry attack, so that thewatermark can not be extracted correctly from it. To solve the problem, a watermark algorithmagainst geometrical attacks based on SIFT is proposed. The proposed takes advantage of theinvariant to rotation, scaling and translation of SIFT, and selects the square of original image’s inscribed circle as the watermarking embedding area; the watermark is embedded in waveletdomain by an optimal quantization of the maximum singular value of each block; at last, the SIFTfeature point is saved as key which is used to detect and correct the geometric attacks. Experimentsresults show that the proposed can resist geometric attacks well such as cropping, rotation, scalingand translation attacks.