The Study And Application Of Neural Network Based Color Image Watermarking

With the rapid development of the digital multimedia technology and the Internet, the spread and exchange of the information , especially the exchange of the digital multimedia , has reached the most intense and broad scope than before . Although the information of the digital multimedia is very easy spreaded and processed, at the same time , it also poses a serious threat to the copyright of the works of the digital media . For example, the ownership of the works of the digital media is infringed ,and the digital works are tampered and that pirated, and so on . Therefore,under the conditions of how to take full advantage of the convenience of the Internet , but also to better protect the copyright of digital media, witch attract much attention. In order to effectively solve this contradiction , the digital watermarking is generated . The digital watermarking technology plays an important role in the current worldwide technology of the protection of the copyright . It will result in some social and economic benefits to a certain extent . So , it is very worth deeply exploring and researching . In this paper , what have been researched to the watermarking technology for the color image are the following :1, Designed a spatial domain digital watermarking algorithm based on the chebyshev neural network and the Zernike . To make the algorithm has the geometric invariance , when watermarks are embedded , these watermarks are embedded in the same domain of the original image with Zernike invariant moment features. To blindly extract these watermarks, also using chebyshev neural network to model , and the neural network can estimate the embedded watermarks. To make these watermarks to be adaptively embedded , the embedded strength can be automatically modified basing on PSNR . Experiments show that the algorithm has better resistance the common noise , JPEG compression , filtering , geometric transformation , image shearing and other attacks than the traditional spatial watermarking algorithm , also has a good transparency .2, Presented a blind digital watermarking algorithm for the color image based on the legendre neural network and Zernike moments Contourlet transform . In order to make the algorithm has the geometric invariance, these watermarks are embedded in the same Contourlet transform domain of the original image with Zernike . To make these watermarks to be blindly extracted, using legendre neural network model to estimate the embedded watermark. The algorithm also can automatically adjust the embedded intensity basing on PSNR . Experiments show that the algorithm can better resist the common noise, JPEG compression , filtering , geometric transformation , image shearing and other attacks , and has strong robustness and great transparency .3, Introduced the application of the digital image watermarking technology in the protection of the copyright of the digital watercolor picture, develops a image watermarking detected system basing on neural network chaos control and asynchronous encryption and NVF with the VC developmental tool , and combining with the characteristics of the digital watercolor picture . The system can effectively embed some characteristic informations of the author and date in the image of the digital watercolor picture , and detect these logo informations . It can better protect the copyright of the digital watercolor picture . The system has the simple implement , stable running and efficiency robustness (an ability to resist shear, rotation, coloring, JPEG compression ) .