Research On Low-Dose CT Projection Domain Noise Reduction And Post-Processing Algorithm Based On Dictionary Learning

Since the first CT machine was successfully developed in 1971,computed tomography(CT)has been widely used in clinical diagnosis after more than 40 years of development.And computed tomography has become one of the important tools in the field of radiation.However,with the popularity of CT imaging technology,the radiation dose of X-ray has been paid more and more attention.Low-dose CT scanning technology,which limits X-ray radiation dose,often reduces the imaging quality of CT images while reducing the radiation dose.How to obtain CT images by improving the imaging technology and using relevant algorithms when the X-ray radiation dose is the lowest.The quality of the CT images obtained is similar to the CT images obtained by standard dose computed tomography,or it is possible to obtain CT images that are more conducive to clinical diagnosis,which naturally becomes an urgent problem in this research field.The content of this article mainly focuses on two aspects of data recovery in low-dose CT projection domain and low-dose CT image quality improvement are based on dictionary learning.The specific content is as follows:(1)Aiming at the noise characteristics of low-dose CT projection data,a maximum posteriori projection domain noise reduction algorithm based on dictionary learning and block-matching 3D filtering is proposed.The algorithm first uses a discriminative dictionary to preprocess low-dose CT projection data to achieve the purpose of suppressing some noise.Then the MAP algorithm framework is used to construct a joint prior model consisting of the median energy function and the BM3 D operator,which smooths the entire projection data,and then improves the quality of low-dose CT images.The experimental results show that there are fewer artifacts in the images reconstructed by the algorithm in this chapter,and theedge information of the images is better maintained.(2)Aiming at the streak artifacts in low-dose CT images,an artifact suppression algorithm based on dictionary learning and equivalent number of looks is proposed.The algorithm first uses stationary wavelet transform to perform single-layer decomposition on low-dose CT images,and then trains the dictionary on the high-frequency parts of the image.Then the dictionary is partitioned by equivalent number of looks to obtain the artifact dictionary and feature dictionary,and only the feature dictionary atoms are sparsely encoded.The processed CT image is obtained after inverse stationary wavelet transform.Then use a bilateral filter to decompose the processed CT image and train a dictionary of high-frequency image.The equivalent number of looks judgment method is used to abandon the artifact dictionary,thereby removing the artifacts and noise remaining in the high-frequency image,and the purpose of suppressing the streak artifacts is achieved.The experimental results show that the algorithm in this chapter retains more edge and detail information while suppressing streak artifacts,and has higher structural similarity and peak signal-to-noise ratio.