Low-Dose CT Reconstruction with Non-Local Functionals

Abstract

In medical CT, the X-ray exposure dose reduction is expected. As a decrease in the dose, the image is degraded due to the noise. Therefore, the development of the noise reduction algorithm while maintaining image quality is an important issue. To suppress the noise, the penalized least squares method is effective. Recently, non-local total variation (NLTV) and non-local structure tensor TV (NLSTV) have been reported. These functional penalties have shown excellent denoising performance of the natural image. In this paper, we apply the functionals to the low-dose CT reconstruction problem. The reconstruction method and the comparison between TV, NLTV, and NLSTV are shown.