Research On Blind Image Watermarking Optimization Algorithm Based On Hybrid Domain

With the continuous development and maturity of information digitization technology and embedded devices,people can not only generate a large amount of multimedia content,but also edit,upload and share these data on the Internet.The availability of Internet data means that content providers must face how to prevent illegal infringement,fraudulent tampering,or any other form of identity verification for sensitive content.Digital watermarking technology is one of the effective methods to solve the above problems,and has been applied in various information security fields.By embedding any assumed content（such as audio,text,data or video）into the original content,it provides authenticity verification and copyright protection for digital content.Embedded data achieves its goal without affecting the quality of the original content.An ideal digital watermarking scheme needs to meet the requirements of digital images in terms of security,robustness and imperceptibility,which has prompted many researchers to work in this direction.In an effort to improve the efficiency and effectiveness of the algorithm,the optimization parameter plays a very important role.However,most existing watermarking algorithms don’t take this into consideration.In this thesis we investigate digital watermarking techniques that optimize the parameters according to the image content,especially explores the robust digital watermarking techniques based on hybrid domain and the human visual system（HVS）.According to the survey results,novel robust watermarking schemes are proposed,which can resist harsh attacks that these watermarking schemes are expected to suffer and achieve high imperceptibility.The main research contents of this thesis are as follows:（1）A blind image watermarking algorithm based on noise visibility function（NVF）is proposed,which uses QR decomposition and discrete cosine transform（DCT）to embed meaningful watermark images into carrier images.In the embedding process,a watermarked image is constructed according to the watermark information and the relationship between the second coefficient and the third coefficient in the first column of orthogonal matrix Q.The optimal embedding strength is obtained based on the balance between watermark embedding energy and perceived transparency.Arnold transform is used to improve the security of the watermarking scheme.In addition,the human visual model based on NVF is used to determine the watermark embedding block.Experimental results show that the proposed scheme has better robustness against various single attacks and combined attacks,and the perceptual quality of the watermarked image outperforms other related methods.（2）Aiming at the local attributes of the image,an optimization algorithm using the firefly algorithm（FA）to search for different embedding strengths is proposed.The proposed algorithm is considered to be a block-based scheme that utilises the variance pixel value for the selection of significant blocks to embed the watermark,which is a binary logo.（1）The first N image blocks with the lowest variance value are selected as the embedding regions（N=embedded watermark bits）.（2）The FA is applied to each target block to search for the best local embedding strength in order to achieve a balance of imperceptibility and robustness.（3）Each watermark bit is embedded by checking the coefficient components U_2,1and U_3,1of the orthogonal matrix U,which is obtained from the singular value decomposition（SVD）of the DCT transformed image block.The extraction process meets the requirements of blind watermarking.（4）For an additional security,the logical chaotic mapping generates a pseudo-random chain,which effectively guarantees the randomness of the positional relationship between the hidden information and the original information.Experimental results show that,compared with some related algorithms,the proposed algorithm not only has higher perception ability,but also provides better or comparable robustness.