Research On Digital Watermarking Based On Local Features Of Color Image

With the rapid development of computer network technology, digital color image has become one of the most important media forms in modern communication because it’s convenient, instant and easy to transport. Products are demonstrated on the website, and people show themselves by uploading photos to their web space. Therefore, how to protect the copyright of color image information has become an urgent solution. Digital watermarking is an effective method for copyright protection. However, the existing digital watermarking algorithms still have many problems to be solved. There is seldom report about the second generation of the color image digital watermarking algorithm. Focusing on the above problems, combining the theory of scale space and color space, this paper focuses on the color images digital watermarking algorithms against geometric attacks. The main research is as follows:(1) The research status of digital image watermarking algorithms against geometric attacks particularly analyzed. Firstly, summarize the main attack methods of digital image watermarking, and the geometric attacks is introduced, and then three kinds of digital image watermarking algorithm against geometric attacks are detailed described. Especially, the typical digital watermark algorithm of the second generation based on image content features is carried on the thorough analysis. Finally, summarize the advantages and disadvantages of the above algorithms.(2) Based on color image invariant feature, a color image watermarking algorithm is proposed. Firstly the original color images are converted from RGB space to YcbCr space. The feature points are extracted by SURF detector based on the component probability density of the color image, and then, according to the characteristic scales, the feature regions are constructed. The watermark embedding and detection are completed in the DFT domain. Finally, the thesis uses a simulation experiment which simulates the real environment of attack by general watermarking testing software Stirmark4.0to test the performance of the algorithm. Experiments have shown that the proposed algorithm is not only nonperceptible but also demonstrates good robustness against conventional signal processing and geometric attacks, especially for local shear. In addition, blind detection algorithm of the watermark is presented. (3) Based on SIFT (Scale Invariant Feature Transform) and color invariance model, a second generation color image watermarking algorithm is proposed. Firstly, the color invariant characteristic variable is extracted by converting the image of RGB color space to color invariance model. The features with color and scale invariance are extracted by the stable SIFT algorithm based on the color invariant characteristic variable, and then, a set of stable and non-overlapped circular regions are chosen by selection criterion based on the clustering algorithm of minimum spanning tree. Main direction normalization is performed on these circular feature regions. Finally, the digital watermark is embedded in the low-pass subband in the NSCT domain, by using odd-even quantization method. Experiments have shown that the proposed algorithm is not only nonperceptible but also demonstrates good robustness against conventional signal processing and geometric attacks, especially for color attacks. In addition, the blind detection algorithm of the watermark is presented.