| Computerized Tomography(CT)imaging technology has developed at full speed being used extensively in the field of clinical diagnosis in recent years.On the other hand,the medical radiation caused by CT scan keeps increasing year by year,and human health is greatly threatened,causing more and more people begin to pay close attention to the problem of the radiation dose.However,the reduction of the dose always comes with the severe deterioration of the CT image quality.Therefore,in the clinical application of low-dose CT,the principle of minimizing radiation dose should be followed in the premise of ensuring high-quality reconstruction images,and the research of low-dose CT imaging technology becomes a hotspot.In this thesis,the projection data restoration method and noisy reconstruction image post-processing methods for low-dose CT are deeply researched in terms of improving the imaging algorithms,the main works are summarized as follows:(1)An algorithm of projection restoration based on residual image decomposition and smooth patch ordering was proposed for low-dose CT.Firstly,the SPO approach was applied to projection filtering for low-dose CT to produce an initial denoised projection.Then,the residual image was decomposed into the structure and the noise parts by the morphological component analysis(MCA)method based on online dictionary learning.The structure part extracted from the residual image,as a compensation image,was added back to the initial denoised projection to generate the restored projection.Finally,the reconstructed image for low-dose CT was obtained by performing filtered-back projection(FBP)from the restored projection.The results from computer experiments demonstrate that the proposed method performs better not only in noise suppression,but also in edge,detail and structure preservation,when compared with traditional reconstruction algorithms for low-dose CT.(2)A novel fractional-order partial differential equation model was proposed for low-dose CT image processing.The traditional Perona-Malik(PM)model has good performance in flat regions,and the total variation(TV)model works better in edge preservation.But these two models often lead to blocky effect.The fractional-order partial differential equation models can mitigate block effect while preserving fine details and more structure.The proposed model is based on the weighted combinations of the fractional-order PM model and the fractional-order TV model,which maintains the advantages of PM model,TV model,and fractional-order partial differential equation models,and overcomes the disadvantages of these three models.Moreover,considering the fine features and details in the neighborhood of the image usually have larger local intensity variance than the noisy background,the local intensity variance is introduced in both weighted coefficient and diffusion coefficient of the proposed model to properly preserve edges and details.Finally,the validity of the proposed algorithm is verified by subjective and objective analysis of the experimental results.(3)An algorithm of modified smooth patch ordering(MSPO)is proposed to improve the low-dose CT images.The distances between the patches were not used and the separate training set was needed in SPO method.In order to overcome these two limitations,the non-local means(NLM)algorithm was modified by replacing the Leclerc robust function with the modified Bisquare robust function,and the modified NLM algorithm was incorporated into the smooth ordering of the pixels scheme,with patch classification and subimage averaging methods.Additionally,before the patch ordering scheme,the prewhitening of the low-dose CT image was carried out to change the correlated noise to uncorrelated noise,in order to remove the noise more easily.After the patch ordering scheme,the TV filter replaced the second iteration method of the SPO algorithm as a post-processing step to further remove the residual noise of the recovered image.The results from computer experiments demonstrate that the proposed approach can suppress steak artifacts and mottle noise,while preserving the important structural information. |
