Research On Image Digital Watermarking Based On Transform Domain

The schemes of traditional cryptography have no ability to protect the copyright of digital products.Firstly,it is easy to draw attackers' attention if these files are encrypted.Secondly,the traditional cryptography can ensure the security of the process when these digital products are transmitted.But it cannot protect these products not to copy and transmit when they are received by users.The digital watermarking technology can solve these problems.The digital watermarking is that the information or image that they can identify owner are embedded into digital products.It should be used to protect copyright,confidential correspondence and so on.There are some problems with it,such as invisibility,robustness,capacity and security.This thesis analyzes the existing schemes,then puts forward two innovative schemes: single and dual watermarking.Firstly,this thesis proposes an algorithm of watermarking pre-process based on PQT and SCAN.It lowers the correlation of spatial pixels and get a high level of scrambling.These bits of error can be distributed throughout the image.In addition,the experimental results show that the proposed method is quite consistent with visual observation based on evaluating indicator of image scrambling.Secondly,this thesis proposes an adaptive watermarking algorithm.The watermark information is embedded into the two layer wavelet transform domain,and the embedding strength is determined by the neural network and genetic algorithm.The neural network is used to classify according to the texture feature of carrier image,such as entropy,contrast,correlation,angular second moment and brightness.Then the best strength of embedding is selected by the genetic algorithm.The experimental results show that this algorithm has better robustness and invisibility.Thirdly,this thesis proposes a self-recovery watermarking algorithm to solve the problem of watermarking capacity.The dual watermarks are embedded into the carrier image and establish a correlation between the first watermark and the second watermark.If one of the watermarks is severely damaged,another can be used to recover.Moreover,extract these points with stable features from the carrier image according to the characteristics of human visual system so that the watermark can be detected when the image is damaged to some extent.The experimental results show that this algorithm has better robustness and invisibility.This thesis proposes an algorithm of watermarking pre-process based on PQT and SCAN,an algorithm of adaptive watermarking algorithm in wavelet domain based on neural network and genetic algorithm and an algorithm of dual self-recovery watermark algorithm to improve imperceptibility,robustness and capacity.This thesis shows the major process and the implement of the strategy.The process of watermark scrambling,embedding and extraction is simulated with MATLAB R2013 b and MyEclipse10 on the local machine and the experimental results show that this algorithm has better robustness and invisibility.