Research On Diffusion Weighted Imaging Denosing

With the development of science and technology,digital images have gradually become a new science and have been widely used in various fields.Its theory and technology will continue to deepen and grow with the continuous exploration of the researchers.This paper mainly studies the denoising method of diffusion-weighted images.Diffusion weighted imaging,as a derivative technology of magnetic resonance imaging,is the only non-invasive brain detection technology,which is widely used in clinical diagnosis and treatment.However,it is often influenced by many factors.In the process of imaging,the noise is introduced,the quality of the image is affected and the quality and readability of the image is reduced.Therefore,the study of the de-noising method for diffusion weighted imaging i s of great significance both in the field of medical application and in the field of digital image.At present,most algorithms assume that noise is known and ignores the diversity of the actual noise.In this paper,the noise of diffusion weighted imaging will be analyzed,and a sufficient theoretical basis is laid for the establishment of the denoising model.The existing de-noising model does not take into account the structural features and image characteristics of the diffusion weighted imaging,and do es not consider the corresponding means for different regions.Aiming at this problem,this paper has made some corresponding improvements based on the non-local means algorithm and the partial differential equation algorithm,and put forward some new idea s,and set up new denoising models.After the model is established,the quality of the restoration needs to be considered.Diffusion-weighted imaging is different from traditional natural images.It is mainly to provide sufficient information for doctor.I n order to improve the correctness of the diagnosis and the effect of the treatment,there is a certain requirement for the preservation of various structural details.The main work of this article is as follows.1.Explain the research status and scientific significance of magnetic resonance image denoising.It introduces the unique imaging characteristics of diffusion-weighted images,the characteristics and sources of noise in images.2.Estimating the noise field of noised diffusion-weighted image.We propose a background-based noise estimation method and a robust iterative algorithm,provides the basis for the construction of subsequent image denoising models.In order to satisfy the particularity of medical image quality assessment,we propose a new i dea for the evaluation of diffusion-weighted image quality base on the existing image quality evaluation standards.3.In this paper,the details of the classical algorithm are given.The disadvantages and advantages of the non-local means denoising algorithm and the partial differential algorithm are analyzed in detail,providing guidance for the improvement of the algorithm and the new algorithm.On the basis of the classical non local mean denoising algorithm,combine with the structure features and the related features of the diffusion-weighted image.4.In order to improve the shortcomings of the original model,the weighted kernel function of Gauss-cosine type and the metric function of symmetry are proposed.The new model not only improves greatly the denoising effect compared with the classical model,but also performs better than the current denoising algorithms based on the non-local mean.5.On the basis of the denoising algorithm based on partial differential equation.It combines the characteristics of the diffusion-weighted image to the Rician noise distribution,and the related information between the images.The adaptive coefficient is constructed,which can distinguish the regional features of pixels well,and the adaptive fidelity term is proposed.According to the feature of image gradient,an adaptive gradient fidelity term is constructed.A double fidelity total variation model is obtained,which not only works well for diffusion-weighted images,but also performs well for traditional images.