The Application Of Neural Network And Binary Operation To Image Watermarking

With the rapid development of computer, network and communication technology, especially the popularization of Internet, the problems of the security protection of information become more and more pressing. The digital watermarking technology as a new information hiding technology, it utilizes the redundant date and random in the digital produces, embeds imperceptible but distinguishable secret messages into the media to protect the copyright or the integrity of the working. Digital watermarking technology is widely studied and used as an effective means of knowledge copyright protection.Firstly, in this dissertation, a novel meaning binary image digital water- marking algorithm based on DWT is proposed. The proposed scheme utilized chaotic sequences generated by general cat map to scramble the watermarks embedding positions, so that the security of embedding watermarks was en- hanced. The improved binary operation methods was utilized,and watermarks are embedded to the low frequency coefficients of deep wavelet domain. Experimental results show that the robustness of this watermarking algorithm is better. Secondly, in this dissertation, a watermarking algorithm based on Full Counter-propagation neural network is proposed. According to the needs of embedding watermarks, we propose a novel method called Full Counter- propagation Neural Network(FCNN) for digital image watermarking. It is complicated for embedding and extracting of trational digital image water- marking. The watermark is embedded in original host image. The watermarked image and watermark image input the neural network to train. In the process of extracting, the watermarked image input the neural network to extract water- mark image. Whether the attacks exist or not, the trained neural network can extract the watermark image. Experiment result for this watermarking algorithmis given. Lastly, a blind digital watermarking algorithm based on Hopfield neural network is proposed. The original host image and original watermark are stored by the Hopfield neural network. In the watermarking embedding, the noise visibility function(NVF) is used for adaptive watermark embedding. In the watermarking extracting, the original host image and the original watermark are retrieved by the neural network.