Research Of Digital Image Watermarking Algorithms

The watermarking which is a effective technology to protect the copyright,integrity of digital multimedia and network security has become one research focus of attention recently. One of the key problems of watermarking is robustness to geometric distortion, which can disable watermark by destroying the synchronization between watermark and carrier image. In this paper, algorithms have been proposed to achieve the robustness of image watermarking against geometric distortion.Combined with the human visual characteristics, the features of SVD against geometric attacks and application are studied, two new digital image watermarking algorithms based on SVD are proposed in the thesis.(1) Adaptive digital watermarking algorithm based on blocking and SVD (Singular Value Decomposition). To solve the single watermark intensity problem of traditional watermarking algorithm, combined with the strong stability of SVD on image, an adaptive digital watermarking algorithm based on blocking-SVD is proposed. The primitive image is divided into blocks in the DWT domain, eigenvalues based on the mean, variance and information entropy are calculated according to the wavelet block's brightness and texture features, then the class selection method is used to select optimal watermark intensity of every image blocks, the singular value of gray watermarking is embedded into the singular value of primitive image. Experimental results show that the proposed algorithm gets over the Blocking Effect problem of traditional algorithm, and has a good transparence and robustness.(2) A blind watermarking algorithm against geometric attacks based on SVD. According to the rotation and scale invariant property of Zernike moments, a blind watermarking algorithm against geometric attacks based on Singular Value quantization (SVD) in DWT domain is proposed in this paper. Firstly, the low-frequency coefficients of the primitive image with DWT are transformed by block-SVD. Then, a binary watermark scrambled by logistic chaotic is embedded by quantizing the singular values of primitive image. Finally, the Zernike moments of primitive image as calibration geometric attacks parameters is used to correct the attacked watermark image, which can strengthen algorithm ability against geometric attacks. The watermark is extracted without the original image. The experiment results show that the algorithm is invisible and strongly robust for the common attacks, and can identify and correct watermark image which have been attacked. The method s performance is much better than the existing and similar watermark algorithms.