Analysis And Research On Robust Zero-watermark Algorithm For Diffusion Tensor Imag

Digital medical images are becoming more widely used in medical diagnosis and clinical treatment as a result of the rapid development and widespread application of Internet technology.As a type of digital carrier,digital medical images,on the one hand,can realize remote diagnosis with the help of the Internet and computer technology,thus effectively improving doctors’ medical diagnosis and meeting the public’s medical and health needs;On the other hand,if digital medical images are not effectively protected by copyright during the transmission process in the Internet,they are extremely vulnerable to the risk of unauthorized copying and malicious tampering,which this not only violates patients’ privacy,but also brings sinister influence to clinical diagnosis,so the copyright protection of digital medical images becomes an important research area of digital medical information security at present.However,most copyright protection algorithms have focused on 2D or 3D medical images,and few algorithms have been proposed for copyright protection of complex medical data,especially Diffusion tensor-Magnetic resonance imaging(DT-MRI)images.DT-MRI images have a five-dimensional data structure and a spatial structure using a tensor representation,so DT-MRI images are characterized by a large amount of data and a high sensitivity to data modification.In addition,in the study of copyright protection of digital medical images,compared with the embedded robust watermarking algorithm,the robust zero watermarking algorithm adopts the form of zero embedding for the host image,which can avoid modifying the original medical image and accomplish copyright protection at the same time.Therefore,the robust zerowatermarking algorithm for DT-MRI images is the focus of this dissertation in order to achieve effective copyright protection for DT-MRI images.The research point of the robust zero-watermarking algorithm based on DT-MRI images is to perform effective feature extraction for DT-MRI images,i.e.,the designed algorithm needs to be able to significantly abstract the high-dimensional features of DTMRI images and the correlations among them,and to ensure the spatio-temporal efficiency of the algorithm;Thus,the proposed zero-watermarking algorithm should be highly robust to common intentional or unintentional image attacks i.e.,the proposed algorithm can still extract the complete zero watermark information or verification information after suffering from various attacks against digital images.To address the special features of DT-MRI images,this dissertation studies and analyzes two different image processing perspectives from the spatial structure characteristics of DT-MRI and tensor data structure features,respectively,and designs and proposes two zerowatermarking algorithms for DT-MRI images with high robustness.The main work in the full dissertation is summarized as follows.(1)Based on the spatial structure characteristics of DT-MRI,an orthogonal raycasting tensor sampling(ORCTS)and quaternion generic polar complex exponential transform(QGPCET)are proposed to provide robust zero-watermarking algorithm for DT-MRI images while ensuring the accuracy and medical validity of DT-MRI image data,and can also greatly improve the efficiency of watermarking.Simulation experimental results show that the proposed algorithm is able to resist common image processing attacks and geometric attacks well,guaranteed watermarking bit correctness rate of1.0000 in most cases.In the best case,the scheme can reduce the watermarking storage consumption to 0.07%.and reduce the time consumption to 0.75%.(2)Based on the tensor data features of DT-MRI images,a tensor image exponential moments(TIEMs)based on the Log-Euclidean measure of Riemannian manifolds is proposed,and TIEMs are also applied to the design of a robust zero-watermarking algorithm for DT-MRI images.The DT-MRI image is transformed and computed exactly on the Log-Euclidean measure based on the Riemann manifold,which is more suitable for the physical meaning and mathematical requirements of the tensor.Simulation experimental results show that the algorithm can effectively resist common image processing attacks and geometric attacks,capable of guaranteeing a bit correctness rate of 1.0000 for watermarking extraction after most attacks,while reducing watermark storage consumption to 11.1% and time consumption to 5.4% compared to traditional zero-watermarking algorithms for 2D medical graphics.