Research On The Digital Image Watermarking Based On HVS In Different Transform Domains

The multimedia network is indispensable source of information in people's lives and works, and the information security and copyright protection of digital products are becoming more and more important. Digital watermarking technology as a supplement to the encryption system, can not only protect the products' copyright, but also can track the spread of digital information and monitor the operation of multimedia content, which has become a focus research area.The dissertation has mainly studied robust digital meaningful watermarking algorithm for still images. Considering the background of digital watermarking technology development and foundational academic frame, according to HVS(Human Visual System) properties, the dissertation proposes two robust digital watermarking algorithms based on different transform domains, which have made good imperceptibility and robustness.Because DWT does not have invariance of displacement, DWT watermark can not effectively resist the rotation attacks, and the sample in DWT also reduces the capacity of the embedding watermark. As to these problems, an undecimated DWT & SVD image digital watermarking algorithm based on HVS is proposed. Because every high-frequency sub-bands have the same amounts of data of the original signals in undecimated DWT, this characteristic can embed greater capacity of watermark data than traditional MALLAT DWT, and SVD watermark is good at Anti-geometric attacking. This algorithm makes use of these advantages to embed watermarks in images. According to HVS characteristics, the dissertation combine texture blocks in space domain and energy blocks in DWT domain to determine the complex texture-energy region. The corresponding sub-block regions in high-frequency sub-band of undecimated DWT are selected to embed gray image watermark with adaptive strength by changing singular-value-size.The limited direction basis function of DWT has made it impossible to describe high-dimensional signals such as images very well. Compared with this demerit, the second-generation Bandelet transform is a novel multi-scale geometric analysis method, which can adaptively trace the geometrical regular surfaces. In this dissertation, the author first reduces the amount of computation and improves the transform speed by ameliorating the geometric angle-direction search methods. According to HVS properties that human eyes are not sensitive to the change of image's textures and edges, the dissertation proposes a digital watermarking algorithm based on HVS in improved Bandelet domain. In this algorithm, meaningful binary watermark information is embedded into the larger transform coefficients which describe these characteristics by odd-even quantization method.Finally, the dissertation concludes the works and points out the direction for further research。...