Research On Robust Digital Image And Video Watermarking Scheme

The rapid development of computer and network technology has greatly facilitated the generation, storage and transmission of digital multimedia data, improved the efficiency of the use of the information, but at the same time, it also brings out intellectual property protection risks. Therefore, it is urgent to need a technology which can effectively protect the rights of individuals. Under this background, the digital watermarking technology highlights the important role.Digital watermarking technology is a very important branch of information hiding technology, in which specific information is embedded into multimedia contents such as images, videos, etc. to protect the property rights. It has been widely applied in many fields. There are many technical requirements for the digital watermarking technology and robustness is one of the most critical factors. However, robust digital image and video watermarking technique have not been researched sufficiently. There are many technical puzzles to be solved, which brings the challenges and opportunities to the researchers.This paper mainly focuses on the key technologies of robust digital image and video watermarking algorithm. Robust digital image watermarking algorithm against signal processing attacks, robust digital image watermarking algorithm against geometric attacks, robust digital video watermarking algorithm against collusion attacks and robust digital video watermarking algorithm against spatio-temporal attacks have been described and researched extensively. In this paper, the research work and the main contributions are as follows:(1) The robust image watermarking algorithm based on the Contourlet Transform and SVD against signal processing attacks is proposed.Contourlet transform has good performance in multi-direction and multi-scale, and can sparsely express the texture and luminance information. Conventional watermarking algorithms in Contourlet domain are not well combined with the internal characteristics of the host, which decreases the overall performance of the algorithm. On the other hand, the SVD (Singular Value Decomposition) of image reflects the inherent stability of content, and does not change significantly with the operation of general signal processing, which is beneficial to enhance the anti-signal processing attacks capability of the watermarking system. According to the energy distribution in the host, texture is classified. As a result, the specific directional subband in Contourlet domain is selected as the watermark embedding region. After the Arnold scrambling, the matrix built by several larger Contourlet coefficients is decomposed by SVD. Combined with HVS (Human Visual System), the adaptive embedding strength factor is adopted and the watermark is embedded into the host by modifying the singular value. Due to the influence of inverse Contourlet transform of the watermark to low-frequency band, the watermark spreads throughout the range from the high-frequency to the low-frequency. The experimental results show that the proposed scheme guarantees the high visual imperceptibility and robustness against the signal processing attacks including filtering, JPEG compression, etc.(2) The robust image watermarking algorithm based on Harris feature points against geometric attacks is proposed.A novel image watermarking algorithm based on Harris feature points in DCT (Discrete Cosine Transform) domain is proposed, due to the idea that the spatial feature points can enhance the robustness of image watermarking scheme. Each detected Harris feature point is considered as the candidate embedded reference, and we search its neighborhoods to determine the feature point with the maximum response value as the local most stable feature point, and generate the square feature region, where the non-overlapping feature region with the maximum response value is selected as the embedded region. Modified odd-even quantization method is applied to embed the watermark into the ZigZag scanned mid-frequency coefficients. The experimental results show that the proposed scheme ensures the high PSNR (Peak Signal Noise Ratio), and is robust against geometric attacks such as rotation, scaling, etc.(3) The robust video watermarking algorithm based on space-time HVS and STDM against collusion attacks is proposed.Collusion attacks are one of the most common types that video watermarking system encounters. Conversional STDM (Spread Transform Dither Modulation) is based on fix-length quantization step, which do not exploit content characteristics of video sufficiently. As a result, the robustness against collusion attacks is relatively weak. Spatio-temporal masking represents visual perceptual redundancies under which human eyes are usually not sensitive. We adopts not only the spatial CSF (contrast sensitivity function), the luminance masking function, the contrast masking function but also the spatio-temporal masking function to account for the model of human eyes to rapid motion and the slow motion in the video, establishes the HVS in the spatio-temporal domain, determines the parameters and the adaptive embedding strength factor and quantization step for STDM. Experimental results show that the proposed scheme guarantees high PSNR and improves the robustness against collusion attacks.(4) The robust video watermarking algorithm based on3-dimensional SIFP against spatio-temporal attacks is proposed.Spatio-temporal feature points represent the great variation in the video, and do not change significantly under spatial and temporal attacks. Therefore, they can be used to improve the robustness of the video watermarking algorithm. The video watermarking scheme based on3-dimensional SIFP (Scale Invariant Feature Points) is proposed. We construct a3-dimensional Gaussian pyramid cube, and in accordance with the extended Hessian matrix, the3-dimensional SIFP detection method is presented, where detected feature points embody not only the scale invariant points, but also reflect the intrinsic characteristics of the significant change in the joint spatio-temporal domain. The feature points are considered as the center to generate the square region. The region is transformed into DCT-SVD domain, and the watermark information is embedded into the segmented mid-frequency coefficients by modifying the norm of the maximum singular value. The experimental results show that the algorithm does not significantly reduce the visual quality, but also is more robust against the variety of attacks, including the frame operations in temporal domain and scaling, JPEG compression in spatial domain.In summary, the target of the dissertation is to improve the robustness of the digital image and video watermarking algorithm. The key issues such as the robustness against signal processing attacks, robustness against geometric attacks, robustness against collusion attacks and robustness against spatio-temporal attacks are researched in detail. By modeling and analyzing of the key problems in robust digital image and video watermarking scheme, corresponding solutions are put forward, and the performance are verified through the simulation results. Based on the above solutions, the robustness in existed watermarking schemes can be improved. The research of this dissertation provides a basis for the further research of robust digital image and video watermarking scheme in the different business requirements environments and plays a positive role in promoting the development and practical application of digital watermarking technology.