2D neural networks for artefact correction during CT reconstruction

Artifacts are often included in actual CT images, sometimes it is due to the nature of the object itself, sometimes it is due to the measuring instrument. The artifacts can seriously affect downstream high-semantic tasks, so it is a basic and important task. In this research, the artifact removal can be seen as a image inpainting or denoising problem. Firstly, it is necessary to characterize the distribution of the artifacts. Secondly, a generative model is planed to be used for artifact removal. Thirdly, the performance of proposed methods will be evaluated using PSNR/SSIM/MS-SSIM metrics. In addition, the proposed image restoration method will be compared on different datasets to demonstrate the generalizability.