The Research Of Geometric Attacks Invariant Digital Image Watermarking Algorithm

In the last ten years,the technology of digital multimedia and internet communication was rapidly developed,especially the popularization of the Internet technology push the development of the digital multimedia application,and have more and more important influence in the people's lives. Many types of digital multimedia can be conveyed in a new way of network,and how to protect the copyright effectively becomes an urgent subject. As the method of protecting the copyright,digital watermarking technology is deeply studied. Resisting geometric attacks is considered as a vital problem to improve the robustness of digital watermarking algorithm, and it is a neck to apply the technology of digital watermarking to the besiness, so it is very important to improve the ability of resisting geometric attacks.This paper takes still images as an example, and the research situation of geometric attacks invariant watermarking algorithm is analyzed firstly, and analyzes the character and shortage. Then it shows the fundamental conception, the property of digital watermarking technology, and the theory of wavelet analysis in digital image processing. After that, the autocorrelation function based geometry transform parameter estimate algorithm is researched.New contributions of this paper are listed as follows:1.Put forward a geometric attacks invariant method embedding the periodical watermark in the discrete wavelet transform domain. Conventional autocorrelation function based watermarking embeds a watermark in the spatial domain, and as a base we designed a geometric attacks invariant using autocorrelation function digital watermarking algorithm. A periodic watermark is embedded in the DWT domain by considering the autocorrelation peak strength and noise visibility. In the process of embedding, the HVS is combined and the watermark is embedded adaptively according to the character of the image.2.Analyze the peaks in the autocorrelation function to estimate the geometric translation parameters of the image.The periodical watermark is used as the template for estimation of the geometric transform parameters because the peaks in the autocorrelation function has regular structure. The rotation angle is estimated from the main axes direction estimated by hough transform becanse the noise peaks is suppressed statistically. The scaling factors is estimated from the distance between the two neighboring lines parallel to the main axes in the peaks. Before detecting the watermark, the geometry transform parameter is estimated, and then the image is renewed.3.Apply the geometric attack invariant testing and the robust testing to the proposed algorithm.Experimental results show that the proposed scheme yields stronger AC peaks than the spatial domain scheme, and shows improved robustness to the common image processing, and estimates the geometry transform parameter of rotation, scaling well. The results show the scheme is robust to compression, and the clarity of image embedded watermark is good.