Iterative limited-angle computed tomography reconstruction method based on diffusion model (iDM) for panel-style industrial assemblies

One of the important applications of limited-angle computed tomography is inspecting chips, printed circuit boards, and other industrial electronic packaging assemblies. Regularization-based methods are relatively mainstream research directions to solve the artifact problem in limited-angle CT reconstruction of these industrial panel-style assemblies', which also could improve the quality of CT images. In practice, it is not easy for the existing iterative methods to meet the requirements of limited-angle CT for both reconstruction quality and fewer iterations. In this paper, we proposed an iterative reconstruction method based on the diffusion model (iDM method) to reduce limited-angle artifact. The method takes full advantage of the features that the gradient and gray values at the visible boundary of the limited-angle CT reconstruction image are both approximately accurate to construct an optimization model. Moreover, the proposed method combines the time-stepping technique and the boundary information to obtain the heat diffusion equation, solved to get high-quality CT images. Numerical experiments show that the proposed method requires only a small number of iterations to obtain the noticeable effect of limited-angle artifact reduction.