Research On Robust Blind Watermarking Algorithm For Diffusion Magnetic Resonance Image

Telemedicine technology is evolving rapidly and now allows for the integrated transmission of voice,images and video for remote diagnosis over high-speed networks.However,the transmission of unprotected medical image data over the network is vulnerable to attacks that can lead to patient privacy leaks and affect physician diagnosis.Therefore,it becomes increasingly urgent to provide accurate and protected medical images.Most of the current copyright protection algorithms are for two-dimensional or three-dimensional medical images,while there are relatively few algorithms for copyright protection of complex,highdimensional medical images.Diffusion MRI(d MRI)is a high-dimensional medical image that can be used to detect diseases,injuries,and anatomical structures in the human nervous system,so providing copyright protection for d MRI images is of high research and application value.In this paper,we propose robust blind watermarking algorithms for diffusion weighted imaging(DWI)and diffusion tensor imaging(DTI),which are common in d MRI images,to improve the robustness of the watermarked signal and ensure the accuracy and clinical value of the images.The main works in the full paper include:(1)A robust blind watermarking algorithm for medical images combining spatial domain features and frequency domain features is proposed to improve the robustness of watermarking signal and ensure the accuracy and clinical value of DWI images.The experimental results show that the algorithm can guarantee the watermark extraction accuracy of 100% in the face of common image attacks,the PSNR of the DWI image after embedding the watermark with the original image reaches 60.06 db,only 5 pixel points with the major axis deflection angle greater than 5°,and the anisotropy score and average dispersion rate are close to0,which meet the clinical diagnosis requirements.(2)A tensor-based robust blind watermarking algorithm is proposed for DTI images.The watermarking algorithm is designed using the transformer framework to ensure the consistency of sampling distance,improve the fusion efficiency,enhance the quality of DTI watermarked images,and combine convolution and transformer to improve the feature space expression capability for more accurate extraction of watermarked signals.Experiments show that the algorithm can guarantee 100% accuracy of watermark extraction in the face of common image attacks,and the PSNR with the original image can reach 50.47 db,which meets the clinical diagnosis requirements.