A novel kernel-based limited-view computerized tomography reconstruction via anisotropic diffusion

In this paper, a novel kernel-based limited-view computed tomography (CT) reconstruction model is proposed. First based on the classical Feldkamp–Davis–Kress (FDK) algorithm, the reconstruction kernel is derived. Then according to the improved approximate inverse for limited view reconstruction, the FDK kernel is truncated to be compact, after which the ring artifacts is reduced. Finally an anisotropic diffusion governed by the Chambolle–Lions equation is introduced to suppress the other artifacts and noise. In the Chambolle–Lions equation, a threshold function is proposed to separate the regions with high or few features adaptively. The efficiency and accuracy of the method are shown in several numerical experiments.

Figure optionsDownload as PowerPoint slideHighlights
► The kernel for FDK algorithm is derived to improve the traditional results.
► Truncate operator is applied to the FDK kernel and the results can be improved.
► PDE based anisotropic diffusion is introduced to reduce artifacts and noise.