The Research Of Digital Watermarking Based On Multiwavelet Transforms

With the rapid development of multimedia and Internet technology, the urgent need tosafety of digital product and information, digital watermarking technology has become a hotproblem in the field of multimedia information security. The study of robust digitalwatermarking against geometric attacks is an important direction in it.Digital watermarking algorithm, based on wavelet transform, has received more and moreattention in recent years. Especially, the appearance of multiwavelet brings greaterimprovement to the research on this field. As the generalization of scalar wavelets,multiwavelets have greater degree of freedom, and have orthogonality, symmetric, compactlysupport and smoothness coincidentally. As a result, multiwavelets have more widely appliedthan scalar wavelets.In this paper, combined multiwavelet transforms with geometric invariants of imagenormalization, we propose a new scheme of image watermarking against on geometric attacks.Firstly, using image normalization and constant centroid theory, we get an interest region withinvariance properties on translation, rotation and scaling. Secondly, the interest region isdecomposed into many sub-images by applying GHM multiwavelet transform. We classifythem into two classes based on the energy proportion for three low-frequency sub-images.Finally, the watermark is adaptively embedded into the two classes through quantizing. weapply different embedding strategies for this two classes, and then use the inverse GHMmultiwavelet transform for all the sub-images to obtain the watermarked image. Theexperimental results show that our algorithm not only has significant robustness on commonsignal processing and geometric attacks, but also a good transparency.By using local image correlation, the algorithm achieves the adaptive regulation of thewatermark embedding strength; it embeds watermark many times, and restores the watermarkinformation based on the method of voting to improve the robustness of the watermarkingalgorithm; it uses the strategy of distortion compensation to compensate the image normalizeddistortion, and enhance the digital watermark imperceptibility.