Design And Implementation Of Robust Multi-watermarking Algorithm For Digital Image

Digital watermarking, which can provide copyright protection and integrality authentication to digital products (video, audio, image and text), is one of the valuable technology to embed some logo information into digital products, and it does not affect the availability of original host data at the same time. With the development of the multimedia and the network technology, the application of the digital products is more widely. Though the way people distribute, get and change information is simple and diversity, the followed problem of copyright protection is more and more serious. The digital watermarking, which can be used to reduce this problem, therefore has become a topic in the academic world. This thesis is focus on the multi-watermarking for images. Through a great deal of experiments, some valuable results are obtained, and the major research work can be summarized as following:1. In this paper, we firstly introduce the background, application and the development of the digital watermarking, and then analysis the related technology of digital watermark generation, embedding and extraction, etc.2. A dual-watermarking algorithm based on DWT and SVD is proposed. Before the watermarks are embedded into the host image, one watermark is firstly embedded into another watermark, and then embedded into the host image. When copyright dispute occurs, we can extract watermark from the intergraded watermark to demonstrate the copyright, and the robustness of the proposed algorithm has been improved.3. An algorithm based on DCT and ICA is proposed, which can resist geometric attacks effectively. We first perform DCT to the host image, divide the host image according to the size of watermark, and then choose the sub-block to embed watermarks combined with HVS. Utilizing ICA, the watermarks can be extracted without original information.4. A robustness method is proposed based on the analysis of the characteristic of the Contourlet transform: (1) embed one watermark into the approximation sub-band of Contourlet transform combined with SVD; (2) select the coefficients with the richest texture from the same coordinate in each finest directional sub-band as the coefficient to embed another watermark; (3) extract the two watermarks from the approximation sub-band and detail sub-band respectively, then perform data fusion to the two extracted watermarks combined with PCA.