Image Reconstruction Algorithms For Sparse View CT

CT radiation dose is closely related to abnormal human metabolism and cancer.At this stage,how to ensure the quality of image reconstruction while reducing CT radiation dose is a huge challenge for CT image reconstruction technology.Among low-dose CT methods,sparse view CT is a recently proposed method to reduce radiation dose by reducing projection views.Because sparse views can reduce radiation dose,reduce scanning workload,dynamic imaging and have a faster reconstruction speed,they have received widespread attention.However,due to the noise and incomplete projection data,the sparse view CT reconstruction problem is a serious ill-posed problem,which leads to severe streak artifacts in the analytical reconstruction algorithm and reduces image quality.In contrast,the iterative reconstruction algorithm based on sparse prior regularization has obvious advantages.It can introduce a priori constraint information according to specific imaging conditions to improve the morbidity of the problem,and ultimately improve the imaging quality of traditional iterative algorithms.Therefore,the research goal of this paper is to solve the ill-posed problem of sparse view CT reconstruction,establish a sparse prior reconstruction model based on regularization to stabilize the sparse view reconstruction,and develop the corresponding iterative optimization algorithm to solve the sparse view CT model,and finally achieve the purpose of improving the quality of reconstructed images in low-dose situations.The main research work of this paper includes the following:(1)To solve the problem that the classic TV-based algorithm is not sensitive to the direction of the image,this paper introduces the direction gradient information of the adaptive image direction,and studies a sparse view CT image reconstruction model with adaptive gradient direction.This model is called BDTV(Block Directional Total Variation)model..In order to solve the BDTV model,this paper adopts an effective gradient descent algorithm,which can effectively update the directionality of the image at each iteration.Experiments show that the algorithm can better preserve the edges of the image and the texture details are clearer.(2)Based on the advantages of wavelet sparse prior information and the ability of TV norm to protect sharp edges of images,this paper improves the model of traditional CT iterative reconstruction algorithm,and studies a new hybrid wavelet and TV two regularized sparse view CT reconstruction model,which can ultimately suppress artifacts and improve image quality.In addition,considering that the two regularization terms produce more parameters,the alternating direction multiplier method(ADMM)algorithm is used to iteratively minimize the minimization problem.The results of simulation experiments and real data experiments show that the method has strong competitiveness in retaining edges,suppressing artifacts and denoising.(3)To solve the bad step effect in the Mumford-Sha-TV model,this paper proposes a Mumford-Shah-TGV model for sparse view CT image reconstruction.This paper uses the elliptic function approximation model of Ambrosio-Tortiroilli under the condition of Gamma convergence,and develops an algorithm based on ADMM iteration.Numerical and real experiments show that the regularization method and algorithm proposed in this paper can reconstruct high-quality images and their edges.