Research Of Application Of Singular Value Decomposition And OSVP In Digital Watermarking Technology

In recent years,with the rapid development of Internet technologies,multimedia products such as images,audio,and video have spread widely on the Internet.Because the multimedia products on the Internet are easy to download and copy,the problem of piracy infringement is increasingly serious.But the appearance of digital watermarks effectively solves the problem of digital copyright caused by piracy.At present,most of the robust digital watermarking algorithms are researched on grayscale images,and algorithms for color host images and color watermarking information are relatively scarce.However,the copyright information used for copyright labeling in life such as icons,trademarks,station signs,and so on,are mostly color images.Therefore,research based on color host image and color watermarks has become a hot topic.In this thesis,a series of research work is done on color host images and color watermarking.The main research contents are as follows:In the research,some problems existing in the traditional SVD watermarking algorithm are found,the solutions and experimental verification are given,and an improved SVD color blind watermarking algorithm based on dual color image watermarking algorithm is proposed.In this thesis,the OSVP model without reference visual evaluation is studied.A new blind watermarking algorithm based on the anti rotation attack of OSVP is proposed,which combines the validity of the OSVP model with the validity of rotation correction and the improved scheme in the previous algorithm.(1)An improved dual color blind watermarking algorithm based on SVD is proposed.The reason of the singular value decomposition theory in the application process of watermarking technology is studied.This thesis,starting from the problem of visual quality deterioration after embedding the watermark,pointed out that the u1v1T matrix after segmentation contains the texture information of the image,u1 prefers the horizontal texture information of the image,and v1 prefers the longitudinal texture of the image and that the traditional algorithm neglects the characteristic of the U matrix as a unitary matrix when modify the coefficient value of the first column of the U matrix,which leads to the serious destruction of the visual quality of the image.And so as to solve the problem of image quality.Focusing on the problem of visual quality destruction of the image in the first column of the U-matrix,this thesis corrects the corresponding algorithm by maintaining the sum of squares of first column equal before and after the modification of the U matrix.Based on the above solutions,the traditional SVD algorithm is improved,and a robust watermarking algorithm with good visual quality and multiple attacks is proposed.(2)A dual color image watermarking algorithm based on OSVP is proposed.To overcome the weak robustness of the rotation attack in the algorithm process,the Orientation Selectivity based Visual Pattern(OSVP)is used.This visual model is mainly used to identify the transformation of the image angle after rotation attack.The image information is corrected by reverse rotation.The OSVP model was originally applied to the image-free reference visual quality assessment.The model has a very good visual quality assessment effect.It can achieve blind extraction in the digital watermark extraction process attributing to the characteristic of non-reference model.In the implementation of the algorithm,the OSVP directional selective visual model and SVD singular value decomposition algorithm are used.Finally,the algorithm is compared with several references from the aspects of invisibility,time efficiency and robustness of the algorithm through simulation experiments.The experimental results show that the algorithm proposed in this thesis is superior in invisibility;For time efficiency,if the rotation correction is performed,the time complexity is high.Conversely,if there is no rotation correction,the time complexity is good.In terms of robustness,it can deal with multiple attacks and has excellent robustness.In this thesis,the suitable object color host image and color watermark image are selected,and the digital watermarking algorithm with good invisibility and robustness is proposed,which can be applied to more real life scenes and is more effective for the copyright protection of digital images.