Robust Digital Image Watermarking Against Geometrical Attacks

With the development of multimedia storage and transmission technology, especially the popularization of the Internet technology pushes the development of the digital media application. Many types of multimedia works such as image, video and music can be conveyed in a new way of network. These technologies bring people many conveniences, but they also bring us some large side effects, the problem of multimedia copyright protection becomes increasingly serious. The protection of the author's copyright becomes an important research topic. As an effective way of copyright protection, watermarking algorithm should be robust against any destroyed attacks. At present, the ability of counteracting attacks, especially geometric attacks, hasn't completely achieved the request of application, it should be improved. So the research on the robust watermarking schemes against geometric attacks is a difficult and hot point.Considering the problems in copyright protection, this paper makes a deep research on the technology of digital image watermarking robust against geometric attacks. Image histogram is a statistical feature, which is resistant to geometric deformations (random removal rows or columns, rotation, scaling et al). The image histogram is almost unchanged after geometric attacks, so watermarking method using histogram could robust to geometric distortions. We propose two watermarking algorithms based on histogram and have perfect experiment results. The main content as follow:First, the paper analyses the background and basic theory of watermarking technology, introduces image watermarking attacks and the methods against the attacks in detail, gives image watermarking algorithms robust against geometric attacks in resent years and discusses their disadvantages. The watermarking algorithm includes exhaustive search, methods using a template, embedding the watermarking in a invariant subspace, methods using feature points of the image, methods using SVD segment and methods using image normalization.Second, with introducing exact histogram specification and image segment, the paper gives a gray image watermarking algorithm based on exact histogram specification. The algorithm considers image histogram as watermark which embeds in the biggest smooth area of the image by exact histogram specification. During embedding process, we restrict the change range of pixels, so the image distortion has little effect on the use of image. The problem of watermark detection becomes simply histogram computation. Experimental results show that the proposed algorithm yields high fidelity degree and relatively high robustness to geometric distortions. It is also shown that this algorithm presents high tolerance to several signal processing approaches including addition of noise, JPEG compression, linear and nonlinear filtering.Third, we propose a color image watermarking algorithm based on histogram feature which modifies B value of CIELAB color model taking advantage of the low sensitivity of the human visual system to changes along the yellow-blue axis. Color image transforms into CIELAB color space, grouping all pixels into blocks, sub-blocks on the basis of their L value which is a mean value of several pixels around the pixel, A value which is a mean value of several pixels around the pixel and B value. We embed watermarking by moving some pixels in a block to form a specified pattern. Experimental results show that the color image distortion is invisible when we embed 64 bits watermarking information. The algorithm is strong robust against geometric attacks and also resist general signal processing including JPEG compression and filtering. The algorithm can extract mostly embedded watermarking information and is a blind detection. At last, we give the summary of this paper.