The Research On Image Quality Improvement Algorithm For Low-dose Computed Tomography

X-ray computed tomography technology develops rapidly in many areas,such as the application of agriculture and forestry,industrial non-destructive detection,material science and medical diagnosis,especially plays a big role in the area of clinical medicine.X-ray radiation can cause a certain degree of harm to the patients and induce cancer and other diseases,so it is the goal of CT researchers to obtain reconstruction image with clear anatomical information and high density resolution while they reduce the CT radiation dose as much as possible.Lowering the tube current is an effective way to reduce the radiation dose,however,this method leads the projection data to produce noise,and then loses the quality of the low-dose CT reconstructed image.This paper uses three methods to remove noise and suppress artifacts,including improving the reconstruction algorithm,projection data filtering,and filtering noise on the reconstruction image.The main innovative work are outlined as follows:1.To overcome the problem of total variation algorithm causing the staircase artifacts and excessive smoothing,an edge indicator function is constructed by combining the weighted variance and image gradient.Combined the diffusion function with the total variation(TV)model,the weighted variance TV model is obtained.Furthermore,the new model was applied to the penalized weighted least square(PWLS),a statistical iterative reconstruction denoising algorithm based on weighted variance TV was presented.To compete the optimal estimation of the new model,two steps were needed.First of all,decompose the joint problems into two sub-problems by alternating direction iterative method.Then,the solutions were resolved by the gradient descent method and the separable paraboloidal surrogates method.Through the visual effects and analysis of quantitative indicators,the new algorithm can significantlyimprove the quality of reconstructed image while obtaining the high resolution of edge details.2.Since the median filter not only can eliminate the impulse noise,but it can also better retain the edge of the image,maximum a posterior projection data filtering algorithm based on median non-local prior was proposed.The new algorithm first focus the median filtering on the projection image,and then performs the adaptive non-local noise reduction according to the similarity of image blocks.The optimal solution of the proposed model is calculated by Gauss-Seidel method.The final reconstruction image is obtained by Filtered Back Projection(FBP).The simulation experiment was carried out by using the modified Shepp-Logan model,showing that the proposed algorithm not only performs well in smoothing projection image noise and suppressing strip artifacts,but also obtains high SNR image.3.The intuition fuzzy entropy(IFE)can adaptively distinguish flat region and the edge detail region of image,and then use it to interact with the diffusion function of the anisotropic diffusion model to propose an edge diffusion function based on IFE.At the same time,a new adaptive total generalized variation(TGV)regularization filter model is obtained by using the new instruction function to improve the TGV model.Finally,the first order primal-dual algorithm was applied to solve the new model and obtain the final reconstruction image.The simulation model and the actual data experiments show that the new algorithm highlights in the noise suppression and strip artifacts reduction while can well preserve the texture characteristics of low-dose CT restoration image.