An Algorithm Of Adaptive Digital Watermarking Based On Wavelet Domain And HVS Color Images

With the development of digital information and computer networks, processing, storage and communication of digital information become more and more convenient and rapid. At the same time, large amount of storage data, lossless copy, unlimited publication of information brought potential threat to the legitimate interests of the owners; meanwhile, it provides a very convenient way to the pirates and others that outlaw. Followed by more and more attention are paid to copyright issues. Digital watermarking technology is an effective solution under this background. In recent years, the emergence of digital watermarking technology is a frontier research area in academia, which is proposed to deal with copyright protection problem in a network environment. The intellectual property of the copyright owner can be protected legally through embedding watermark information into the carrier. The digital watermark embedded into digital works should have at the following characteristics:provable, un-perceptibility, robustness.Digital watermarking algorithm based on grayscale image has become a research focus which has been widely studied and applied. But in the real life, the application of color image occupies the dominant position. Therefore, the focus is turned to digital watermarking technology of color image gradually in recent years. Most of gray image watermarking algorithm can be extended to color images easily. Some watermarking algorithm for color image treat the RGB color image as a "heap", which is made up with three unrelated gray image, and watermark is embedded into the three components or one of them.In the paper, a digital watermarking algorithm of color image based on DWT is proposed, which is combined with HVS. The YCbCr color model is used in the algorithm. The luminance component Y in original color image is picked up and transformed into wavelet domain. Then, the JND threshold model is created according to the cover up factor of the frequency, the sensitivity to the noise in brightness area darkness area of the image, the sensitivity to the noise of image texture complexity area for human eye. And the threshold is normalized through adding an expansion factor, so that the value is less than 1 after normalized. Grayscale watermark image is scrambled as Arnold method. Then the watermark image after scrambled ia transformed into wavelet domain. To ensure the security of the watermark image the iteration time is used as a key. The horizontal, vertical, diagonal wavelet coefficients of luminance component Y is extracted arranged according to the size of each coefficient and subtracted. And then the wavelet coefficients of watermark image are embedded into the wavelet coefficients of luminance according to the subtracting value. Finally, the coefficients of the luminance component Y with watermarking are executed with inverse wavelet transformation. Lastly, reconstructs with the original image in the Cb and Cr components, the color image embedded with watermarking is obtained. Experiment shows that the digital watermark model can effectively guarantee the invisibility; and the improved JND threshold can effectively guarantee the robustness of the watermark and resist various attacks against the color image.However, in the paper, the method also has some shortcomings. For example, it needs to use the original image to extract the watermark; and the robust of algorithm is weak to geometric attacks; If the image is added larg noise, the wavelet coefficients will be changed largely after transformation. Due to the current research on color image digital watermarking algorithm is not enough, the color digital watermark algorithm based on wavelet transformation still has a large optimization space.