Research On Robust Watermarking Algorithm In Wavelet Domain Based On Matrix Decomposition

In the era of Internet prevalence,people can obtain digital media resources conveniently and quickly.But at the same time,copyright problem has arisen,and the management and protection of digital content has received increasing attention.As the most effective way to protect the copyright of digital works,digital watermarking has a wide range of application prospects.In the copyright protection of digital image,robust invisible watermarking plays an extremely important role.In order to ensure the robustness of the watermark,the watermarking algorithm usually chooses to embed the watermark in the transform domain of host image.Matrix decomposition method has an important application in image transformation.If it is directly used for the decomposition of host image,the performance of watermarking algorithm is poor.However,the combination of matrix decomposition and wavelet transform can improve the performance of the watermarking algorithm.Most of the existing watermarking algorithms combine wavelet transform with singular value decomposition.Although the performance of the algorithm is improved to a great extent,there are still the following three problems:1)the reliability of watermark is not considered in the algorithm;2)the wavelet domain watermarking algorithm has poor resistance to geometric attacks,and the watermark capacity needs to be improved;3)the security of watermark has not been paid enough attention.In view of the above problems,taking advantage of matrix factorization,and combining with the proposed new wavelet transform,we have given some reliable and robust watermarking algorithms against geometric attacks,and highly secure image encryption schemes.The details are as follows:(1)Three schemes are proposed to ensure the reliability of the algorithm:First,encrypt the watermark image before performing singular value decomposition on it.Second,calculate the right principal component of the watermark after applying singular value decomposition to the watermark and add it to the host image.Third,propose the random multiresolution Hessenberg decomposition,and use it in the decomposition of watermark.Simulation results show that these three schemes ensure the reliability of the watermark,and the algorithm has no the false positive problem.(2)Five robust image watermarking algorithms are proposed:Based on the good stability of singular value decomposition,three gray image watermarking algorithms are given.First,a hybrid domain watermarking algorithm is proposed.This algorithm uses the generalized Arnold transform to encrypt the watermark to ensure its reliability.Moreover,the embedding matrix is used instead of the single embedding factor to improve the robustness of the algorithm.Second,a double encryption watermarking algorithm is proposed in the invariant wavelet domain.This algorithm uses the generalized Arnold transform and fractional Fourier transform to double encrypt the enlarged watermark to solve the false positive problem.Moreover,the value of the embedding factor is determined by the proposed multi-parameter particle swarm optimization algorithm.Third,a robust watermarking algorithm with large capacity and high security is presented.In the embedding algorithm,the guided dynamic particle swarm optimization algorithm is used to optimize two embedding factors.Combining QR decomposition and the proposed redistributed invariant integer wavelet transform,an invisible watermarking algorithm with high robustness is designed.In the watermarking algorithm,the proposed two-level encryption scheme is used to encrypt the information generated in the embedding process,which avoids the false positive problem and ensures the security of the watermark.Based on the proposed offset linear canonical redundant wavelet packet transform and random multiresolution Hessenberg decomposition,a double color image watermarking algorithm is presented.In the embedding process,the high-frequency part of the watermark is added to the sub-band of the corresponding component in the host image,which improves the watermark capacity and robustness of the algorithm.The simulation and comparative experiments show that the five proposed watermarking algorithms meet the four requirements of robust watermarking.Moreover,compared with the related algorithms,our algorithm has greater advantages,especially in robustness.(3)Two encryption schemes with high security are proposed:First,we present a two-level encryption scheme based on chaos mapping,QR decomposition and Gyrator transform.Second,the proposed random multiresolution Hessenberg decomposition is used to encrypt the image.The simulation results show that the proposed two encryption schemes have high security.