Research On Digital Watermarking Of Static Image

As the development of the internet and multimedia technique, we can obtain information expediently from internet. At the same time, we have the problem - our information can be attacked easily and digital productions will be robbed easily. The copyright can't be protected. For the well-development of internet, the copyright of digital productions should be protected. And the digital watermarking is working for that. The copyright can be identified by watermark embedded in the production reconditely.There are so many kinds of watermark. In this article, we introduce two watermarks based on chaos - Logistic map and toral automorphisms map. The toral automorphisms map is periodic, which is used for scrambling image usually. In this article, we give the inv-map of toral automorphisms which can improve inv-scrambling speed. We also give a image scrambled scheme based on affine transformation, which have long period.The watermarking schemes are so many too. The artical introduces two methods based on DCT and a method based on quantizing the biggest singular value. We also give a scheme in terms of quantizing block-means:(1) Discrete cosine transformation (DCT) is a classical image transformation, which can focus image information and divide the image frequency into three bands - low, middle, high. So it can be used for embedding watermark. We embed watermark in low-frequency coefficient and DC coefficient respectively. The experimental results show that the scheme of embedding watermark in DC coefficient has high robustness. But the scheme of embedding watermark in low frequency coefficient is fragile because some of low frequency is too small.(2) The property of matrix singular value decomposition is similar to DCT. It collects the most energy of image on the biggest singular value which is stable when the image be attacked. Embedding watermark in the biggest singular value by quantizing it. The experimental results show that the scheme is a publishable robust watermarking scheme.(3) We analysis the stability of block-means by image attacking experiment, and find that the block-means are not change much when image attacked. So it is able...