Robust Watermarking Algorithms For Medical Image Based On Transform Domain And Perceptual Hashing

With the development of smart healthcare,the electronic medical information generated by the hospital every day grows exponentially,and medical image information accounts for more than 80% of the entire medical information.When such information is transmitted,stored and shared on the network,it is very easy to cause the leakage of the personal information of patients stored in medical images and pose security risks.Digital watermarking technology is an effective way of protection.However,compared with natural images,medical images have different storage modes,coding structure,information load,etc.,and they have special requirements to assist doctors in diagnosis.Generally,they are not allowed to be modified in order to avoid misdiagnosis.Therefore,it is impossible to directly transplant relatively mature digital watermarking technology to medical image watermarking.In order to avoid affecting the image quality,the existing medical image watermarking technology generally embeds watermark into the region of non-interest of the medical image.This is a kind of spatial domain method,which not only increases the amount of calculation and affects the visual quality of the image,but also limits the embedding capacity of the watermark.The capacity,invisibility,robustness and security of digital watermarking restrict each other,and the existing algorithms cannot resist geometric attacks effectively.In the medical field,this is also the difficulty of research.Moreover,with the rapid development of 5G communication technology,artificial intelligence technology,big data,cloud computing and cloud storage technology,traditional two-dimensional medical images are more presented in the form of medical volume data.These threedimensional data have different features from two-dimensional medical images and are the development direction of medical images in the future,but there are few relevant researches at present.To solve the above problems,in this paper,based on the features of transform domain and perceptual hash,four kinds of perceptual hash models are designed.Combining of human vision system,chaos,homomorphic encryption,zero watermarking and the "third party" concept,several key issues of watermarking in medical image are studied.(1)Aiming at the special requirements of invisibility and robustness of medical images,a robust watermarking algorithm for medical images based on the mean perceptual hash of NSCT-RDWT-DCT is proposed.By introducing NSCT,RDWT,DCT and perceptual hashing,the algorithm model of NSCT-RDWT-DCT is designed,and the robustness and uniqueness of the new perceptual hashing algorithm are proved.It uses NSCT to abandon the sampling step,has a better feature extraction rate,can well resist geometric attacks;uses RDWT to reduce redundancy,can effectively eliminate noise;and uses low-frequency of DCT anti-compression ability and good robustness advantages.It overcomes the visual redundancy of traditional perceptual hashing,improves the calculation accuracy and the defect that it cannot resist geometric attacks.This feature is applied to the design of two-dimensional watermarking scheme,combined with Cat chaotic encryption watermarking.The proposed algorithm is compared with the existing four state-of-the-art algorithms.It has strong robustness and good invisibility,and can effectively protect the personal information in the medical images of patients.(2)On this basis,for the problem that medical watermarking cannot resist geometric attacks effectively,a anti-geometric attack watermarking algorithm based on SIFT-DCT subtract perceptual hash is proposed.Combined with the requirements,the henon chaotic encryption algorithm is designed first,and then based on the analysis of the watermark features of each transform domain and the proposed transform domain algorithm,the subtract perceptual hash algorithm using SIFT-DCT is proposed.The algorithm model is designed,the geometric resistance is tested,and the geometric watermarking algorithm is designed.In the experimental part,the geometrical transformations such as rotation attacks,move down,move left,zoom out,zoom in and cropping attacks are tested,analyzed and compared with algorithms in detail.The simulation results show that the proposed algorithm has good anti-geometric attacks performance.(3)Based on the two-dimensional medical image watermarking and the features of the three-dimensional medical image,a large-scale watermarking algorithm for 3D medical images based on the 3D DTCWT-DCT difference perceptual hash is proposed.After analyzing the transform domain,the DTCWT transform with finite redundancy,better translation and directivity,and high efficiency order is selected to combine with the DCT transform with good compression resistance and robustness.In its transform matrix,combining with human visual features,a difference perceptual hash method is proposed.It can be bound to the robust feature of large-capacity watermarking,and the embedding and extraction method of large-capacity watermarking is designed accordingly.The proposed 3D medical image large-scale watermarking algorithm based on DTCWT-DCT difference perceptual hashing can be adaptive according to the watermarking capacity,without the limitation of embedding capacity,and can effectively resist various attacks while ensuring diagnostic accuracy.(4)It is difficult to analyze and diagnose diseases based on medical volume data due to the large amount of data.If it is transmitted in the network,due to the need for real-time diagnosis,the transmission speed cannot be sacrificed,the security of transmission and the robustness of the watermark must be guaranteed,a zero watermarking algorithm for encrypted medical volume data based on FDCT multi-layer perceptual hashing is proposed.According to the design of the human visual perception model combined with the features of the transform domain data,the concepts of the Chen hyperchaotic scrambling algorithm,three-dimensional FDCT,multi-layer perceptual hashing and cryptography are used comprehensively.The fusion features of medical volume data are used to embed and extract watermark information.The algorithm is homomorphic,and the extraction of watermarks can be completed without decrypting the medical volume data.The designed FDCT multi-layer perceptual hash generates small amount of feature data,which facilitates network transmission and storage to meet the requirements of security and bandwidth.Meanwhile,the introduction of zero watermarking technology enables medical volume data to always maintain high quality,effectively assists doctors in diagnosis,and the encryption algorithm for medical volume data is reversible.The simulation results show that the proposed algorithm makes up for the disadvantages of the existing algorithms such as poor security,robustness and large computation,and can better resist common attacks and geometric attacks.With the advantages of high security,low complexity and strong robustness,it can better solve the storage and transmission problems of massive medical data,and has a broad application prospect in the field of intelligent medical treatment in the future.