Fast de-streaking method using plain neural network

Radon transform fundamentally underlies reconstructions from computed tomography. Radon transform results in a sinogram where each coefficient represents the integral of the image on one line from one angle. From a full Radon sinogram, one can rebuild the initial image with Filtered Back Projection (FBP) algorithm. However, when information from some angles are missing, streaks appear on the output image using such method. In this work, we propose a method that works exclusively on the image domain: a de-streaker based on plain neural network. The objective is not to replace reconstruction methods, but to offer a fast post-processing that reduces artifacts from the output image, particularly when sinogram is unavailable.