摘要
Purpose:Sparse-view imaging is a promising scanning strategy for computed tomography(CT).Conventional total variation(TV) regularization makes use of the sparsity of image gradient.However,conventional TV method may result blocky artifacts in images.Image curvature is one of the second order information and total absolute curvature(TAC) can serve as a feasible supplement of conventional TV.This paper proposes a TAC-TV reconstruction model and its efficient algorithm which combines both first and second order information.Method:The TAC induced penalty function has a better description on the gradually varying images compared with TV-based ones.Consequently,incorporating the TAC for image reconstruction,the reconstruction problem is built as a constrained optimization.The objective function is formed as the minimization of TAC and TV with the constraints of data observation.Both first and second order information is encoded in the optimization.In the algorithm design,in order to obtain an efficient implementation,the alternating direction method(ADM) is utilized in the algorithm derivation which provides stable and fast convergence property.Results:A series of experiments including the inverse-crime studies,noiseless data and real CT data are conducted to investigate the proposed algorithm from sparse-view of projections.Experimental results validate the proposed method has the potential to outperform conventional TV-based reconstruction methods in preserving edge and fine structures.Conclusion:We introduce a new TAC-TV model for CT image reconstruction from sparse-view data and provide an efficient and fast numerical algorithm based on the augmented Lagrangian method.The experimental studies suggest that TAC-TV model can reconstruct higher accuracy images with sparse-view CT.
Purpose:Sparse-view imaging is a promising scanning strategy for computed tomography(CT).Conventional total variation(TV) regularization makes use of the sparsity of image gradient.However,conventional TV method may result blocky artifacts in images.Image curvature is one of the second order information and total absolute curvature(TAC) can serve as a feasible supplement of conventional TV.This paper proposes a TAC-TV reconstruction model and its efficient algorithm which combines both first and second order information.Method:The TAC induced penalty function has a better description on the gradually varying images compared with TV-based ones.Consequently,incorporating the TAC for image reconstruction,the reconstruction problem is built as a constrained optimization.The objective function is formed as the minimization of TAC and TV with the constraints of data observation.Both first and second order information is encoded in the optimization.In the algorithm design,in order to obtain an efficient implementation,the alternating direction method(ADM) is utilized in the algorithm derivation which provides stable and fast convergence property.Results:A series of experiments including the inverse-crime studies,noiseless data and real CT data are conducted to investigate the proposed algorithm from sparse-view of projections.Experimental results validate the proposed method has the potential to outperform conventional TV-based reconstruction methods in preserving edge and fine structures.Conclusion:We introduce a new TAC-TV model for CT image reconstruction from sparse-view data and provide an efficient and fast numerical algorithm based on the augmented Lagrangian method.The experimental studies suggest that TAC-TV model can reconstruct higher accuracy images with sparse-view CT.
引文