Strong Robustness Zero Watermarking Based On Low Frequency Means

With the popularization of network,security problems such as information tampering and embezzlement are increasing gradually,and digital watermarking technology can realize the protection of information security.Traditional watermarking algorithms can not improve robustness and transparency at the same time.It's necessary to confirm the best embedding coefficient by iteration test,and the robustness of geometric attacks is poor.Some algorithms have high false alarm rate.To solve the above problems,a zero-watermarking robust algorithm based on low-frequency mean is proposed.The aim is to improve the robustness of the algorithm while taking into account the false alarm rate and transparency.Through the selection of feature regions and the selection of construction methods,a zero-watermarking algorithm with strong robustness and low false alarm rate is realized.Firstly,the carrier image is transformed by discrete wavelet transform,and the low-frequency region is transformed by discrete shearlet transform,and the low-frequency components in the optimal scale are selected as the stable feature regions.Secondly,the feature region is transformed by block discrete cosine transform,and the eigenvalue moment is Obtained by comparing the size relationship between the mean of each singular value and the mean of the global singular value.Lastly,Zero-watermarking is obtained by XOR operation between eigenvalue matrix and Arnold scrambled binary watermarking.The algorithm adopts the idea of double construction.Selecting two different regions of center and upper left corner to Construct zero-watermarking,so that the high integrity region can be selected for subsequent processing after being attacked by shearing.For rotation attack,the carrier image is prejudged before extracting the watermarking.If attacked,Radon transform is used to correct the carrier image,then the watermarking information is extracted to improve the ability of resisting geometric attack.The experimental results on different texture images show that the proposed algorithm is robust to various attacks and has a low false alarm rate.In addition,because the algorithm does not change the carrier image,it can guarantee the transparency of the carrier image.This thesis has 21 figures,16 tables and 51 references.