Design And Research Of Several Digital Watermarking Methods

Digital watermark technique, as an important tool for copyright protection and information authentication of multimedia, has been widely studied by many scholars, and gat many excellent achievements. But there are still some problems that haven’t been solved, such as low quality of restored image or less robustness. To solve these problems, this thesis has researched the fragile watermarking and robust watermarking techniques and designed several digital watermarking schemes. The main contents are as follows:(1) A novel image recovery method based on Discrete Cosine Transformation (DCT) and matched blocks is presented in this section. This method applies finding the best matched blocks by using a part of fractal compression algorithm for image recovery. At the same time, a new check algorithm for checking if image blocks are tampered is proposed. The best matched block of each small block in the original image is searched in a particular way. Then the matching information is embedded as watermark into other blocks. For the ones that fail to find the best matched blocks, DCT is applied on them and then quantized to be as the reference watermark. Experimental results show that the proposed method has better restorability than the methods in other references. And the lower the tampering rate is, the better quality of restored content can be obtained.(2) Authentication and recovery of images based on standard deviation (STD) is proposed. A better scheme for image authentication and restoration using two watermarks is applied, the authentication watermark for detecting tampered regions, and the recovery watermark for reconstructing reference image to recover the tampered content. The authentication watermark is the binary sequence that is transformed from the difference of the original image and chaotic sequence produced by the Piecewise Linear Chaotic Map system (PWLCM). The reference watermark is compressed from three types that are classified based on the STD of every block sorted in ascending order of quantization DCT coefficients, in which blocks corresponding to different types are encoded to variable lengths. The proposed method obviously improves the quality of reconstructed images.(3) Robust watermark scheme based on discrete wavelet transform (DWT), Singular Value Decomposition (SVD) and Neural Network is presented. The watermarking information is embedded into the singular matrix of the cover image in the wavelet domain. After extracted, use the associative memory network of generalized inverse matrix to recover the extracted watermark. The scheme is robust against various attacks and the quality of the retrieved watermark image recovered by neural network can be comparable to that of the original watermark.(4) A fast fractal encoding algorithm using standard deviation and DCT coefficients is proposed. First, the STD values of domain blocks are calculated. Then, domain blocks are sorted by their STD values. During the encoding process, each range block is limited to search in the domain blocks with similar STD values. Since the searching space is reduced, the encoding speed is improved. Moreover, for improving the quality of the retrieved image, an auxiliary encoding algorithm is proposed, and it could encode range blocks in which the pixel values fluctuate greatly. In the decoding step, a kind of simple line fitting on block boundaries is used to reduce the blocking effects. Simulation results show that the runtime of the proposed method is reduced greatly compared to the full search method. And the new algorithm could obtain good quality and higher compression ratio than the baseline encoding algorithm.