Research On Denoising Of Diffusion Weighted Image

Diffusion-weighted imaging is a non-invasive method for detecting the diffusion of water molecules in living tissue.It is a new magnetic functional imaging technology with strong practicality in the medical field.However,as the diffusion-weighted image is inevitably interfered by noise in the process of acquisition and transmission,and the diffusion-weighted image is sensitive to noise.So it will affect the accuracy of the diffusion-weighted image,which in turn affects the subsequent processing.Therefore,it is very important to reduce the noise in diffusion-weighted image.Diffusionweighted images mainly have two characteristics,one is that the image has a high degree of self-similarity,the texture details are rich,and the other is that the noise contained in the image is Rician noise.In this study,three denoising algorithms are proposed to denoise diffusion-weighted images based on the characteristics of the images,aiming to preserve the texture details and at the same time achieve better noise reduction effect.Firstly,the weighted nuclear norm denoising algorithm is applied to the diffusion-weighted images denoising.The algorithm utilizes the similarity between non-local blocks to denoise the image.The algorithm is relatively novel and has not been used in the current research on diffusion-weighted image denoising.Secondly,a weighted kernel nonlocal means filtering algorithm is proposed in the paper.The Euclidean distance is added to the kernel function of the traditional non-local means filtering algorithm,and then assign weights to neighboring pixels.The algorithm makes the image similarity judgment more accurate and the image can retain more details.Thirdly,a preprocessing method for Rician correction is proposed in this paper.The Rician noise in the diffusion-weighted image is processed by Rician correction first,and then the corrected image is denoised by the denoising algorithm.In order to verify the effectiveness of the above algorithm on the diffusion-weighted image denoising,this paper uses a variety of algorithms commonly used for diffusion-weighted image denoising as a comparison algorithm.In this study,the experiment was divided into three parts: simulation data,fibercup data set and acquisition data.And the experiments of the three parts were quantitatively analyzed,and various algorithms were compared and analyzed in detail.At the same time,the three proposed denoising algorithms are also compared and analyzed correspondingly.The experimental results show that the above various algorithms can effectively reduce the noise in the diffusion-weighted image,and can retain more texture details.