A new weighted anisotropic total variation algorithm for limited angle tomography

Abstract

This paper addresses streak reduction in limited angle tomography. The weighted total variation (wTV) algorithm is able to remove most Poisson noise and small streaks for clinical limited angle data. However, it is still not sufficient to remove larger streaks whose orientations are mostly dependent on the scan trajectory. We propose a new weighted anisotropic total variation (waTV) algorithm, which uses four neighboring pixels to calculate the gradient along the streaks' normal direction and is able to reduce large streaks well.