This repository contains the PyTorch implementation of the paper: 3-D Computed Laminography based on a Sequential Regularization
Abstract: Accurate reconstruction of computed laminography (CL) is a challenging task because projections from the CL scan are incomplete, resulting in inter-slice aliasing and blurring. Hence, we establish an effective CL reconstruction model based on sequential regularization (SR) terms, which include the norm of gradient along different directions and the truncated adaptive-weighted total variation (TAwTV). They facilitate edge-preserving diffusion and enhance the smoothness of the reconstructed object. To solve the proposed model, we introduce an alternating minimization algorithm that decomposes it into several sub-problems, which are solved by the Split-Bregman frame and gradient descent method. Compared with several iterative reconstruction methods, the experimental results demonstrate the effectiveness of the proposed method in terms of preserving edges, suppressing inter-slice aliasing, and reducing noise.
- SART-SR can be installed from source,
git clone https://github.com/YuhangLiu98/SART-SR.git- Then, TIGRE is required, for example,
git clone https://github.com/CERN/TIGRE.git-
Then, the three files for TAwTV need to be placed in the corresponding folders, and compilation information for TAwTV should be added to the 'compile.m' file. Refer to TIGRE's TV for guidance.
-
Lastly, place "SART-SR" code in the directory of "TIGRE":
.\TIGRE-master\MATLAB\Demos\run `PCB_POCS_L0.m` to reconstruct the phantom.
