Color Image Watermark The Nonlinear Detection Program Studies

The rapid development of digital technology and internet technology makes multi-media works (Image, Video and Audio) grow quickly and prevail extensively, but the piracy becomes increasingly severe. Thus, in order to safeguard the lawful rights of creators, the protection of the copyright of digital products is a key problem for the flourish of cultural market. In this situation, the digital watermarking technology was proposed. The digital watermarking technology proves the ownership of multimedia works by embedding secret information, i.e. watermark, into the original digital products. Since the digital watermarking has such significant features as invisibility, non-detection, robustness, and security, it becomes a powerful tool for protecting the copyrights of multi-media works.In this thesis, we mainly study the ccolor image watermarking. We improve the method of Barni's algorithm of the discrete cosine transformation (DCT)-domain images watermarking. In the Red-Green-Blue (RGB) system of color images, the R, G and B bands are extracted respectively, which corresponds to different color matrix. Then, by selecting one band of RGB, the DCT of this band is computed. We resort the DCT coefficients by their values, and view it as a Bernoulli distribution. In this way, the classical multiplicative algorithm of Barni becomes an additional one. Then, we deduce a nonlinear watermarking detector by the weak-signal detection theory in additive noise. Finally, we analyze the statistical characteristics of this detector and test the robustness of this algorithm under various attacks.