On Digital Watermarking Algorithm Research Against Geometric Attack

Digital watermarking as an effective way for multimedia information authentication and copyright protection, has been the hot research topic in the multimedia information security field.The theory and method of static digital image watermark have been studied and applied in this dissertation. The major works involved are as follows:1,In this thesis, the recent development, universal model, classification, characteristics, evaluation criteria, application and representative algorithms of the digital watermark are summarized.2,An improved general Arnold map is used to scramble image pixels in the spatial-domain, and the sequence values of an improved Lorenz chaotic system is adopted to generate keys for the second grayscale image encryption. The key space is large enough to avoid the deciphering by the phase-space reconstruction theory and exhaustive method. The experiment demonstrates the algorithm is effective in the image encryption.3,Based on the large sequence values feature of the Multidimensional chaotic system, a novel algorithm only uses the hyperchaotic Liu system to scramble and encrypt the pixels of arbitrary size color image in the YCbCr-domain. Experiment result demonstrates the algorithm is effective in the image encryption. The ciphered image has random-like distribution behavior of YCbCr values, adjacent pixels have zero co-correlation properties and the deciphered image is perfect.4,Through combination of spatial and frequency domain, the binary watermark is embedded into the colorful image. According to the particularity of the R,G,B three components, the algorithm chooses the B component spatial domain to embed the watermark in order to enforce the watermark imperceptibility, and chooses the G component frequency domain to embed the watermark in order to enforce the watermark anti-compression. The two embed ways are all blind extraction. Moreover, the algorithm can satisfy the current need for the watermark imperceptibility and robustness by means of adjusting to the quantization parameter and block parameter. Experiment result demonstrates the algorithm can effectively resist large degree of geometric attacks and compound attacks, and it is also strongly robust against common image processing attacks.5,A novel geometric attack resistant zero-watermark algorithm based on frequency domain is proposed. The original image and watermark are divided into blocks firstly, and then each block is transformed using DCT and SVD. The method of dividing into blocks enhance secret of embedding watermark, anti-cutting and anti-compression. Using the feature of the singular value sequence, the algorithm achieves embedding zero-watermark and blind watermark extraction. Experiment result demonstrates the algorithm is strongly robust.In the end, the main research work is summarized. The problems exist in the work and the direction for the further study is indicated.