Low-Dose CT Image Denoising Based On Moving Decomposition Framework And Shearlet Transform

Medical computerized tomography(CT)images have become an important tool for modern medical diagnosis and evaluation.The CT imaging is a mature technology,the quality of CT images can reach a very clear level when the scanning dose reaches a certain extent.However,patients will be exposed to strong radiation exposure to CT,thus negative affecting the health of patients.Low-dose CT can significantly reduce the amount of radiation examined by the original CT.However,reducing the X-ray current will lead to the quality degradation of the reconstructed image.Therefore,the denoising of low-dose CT images is of great significance.The research in this paper focuses on the denoising of low-dose CT images.After studying the signal distribution of noise in low-dose CT images and the existing denoising methods of low-dose CT images,in this paper,a low-dose CT image denoising algorithm based on shearlet transform and moving decomposition framework is proposed.In recent years,the research of application of multi-scale transformation in image signal processing has been the hot spot scholar's research.Shearlet transform has more advantages than wavelet transform,such as shearlets has more approximation direction to the image signal,therefore,it has better sparse approximation ability than wavelet transform.The ideal situation of the process of image denoising is that the original details of the image will not be destroyed while the noise is removed,thus a number of image denoising algorithms try to separate the details of the image from the low frequency part of the image.In recent years,a decomposition framework that can decompose the detail and non-detail parts of the image from the original image has been proposed.That framework encodes the pixel-by-pixel movement of the image,then decomposes the origin image into two parts,so that decomposition algorithms is called moving decomposition framework(MDF).The proposed denoising algorithm combines the shearlet transform and decomposition framework which was mentioned above.Firstly,the algorithm decomposes the image into two parts,which contains the image detail and the image without the detail information.The shearlet transform based denoising and block matching 3D-filtering(BM3D)denoising algorithms are then used for each aspect of two parts.Thirdly,the low frequency filtering is applied to the low frequency image obtained by shearlet transform.Finally,the final denoised image was obtained by applying the inverse transformation method of the decomposition framework to two de-noised components.The noisy image is decomposed by the decomposition frame and processed respectively so that the details and edges of the image can be better preserved.The multi-directivity of shearlet makes it get better results in denoising the components of detail.BM3 D as the hybrid domain denoising algorithm can also obtain ideal denoising results when denoising the approximate components of the image,the results can not only ensure the effective of image denoising,but also preserve the details of the image.In the experimental part of this paper,the effectiveness and superiority of this algorithm for low-dose CT image de-noising can be proved by comparing with the existing image de-noising algorithm.