Artificial neural network enhanced total generalized variation regularization few-view CT image reconstruction

X-ray radiation is harmful to human health. So how to obtain a better reconstructed image with few-view constraints is a major challenge in the computed tomography (CT) field. We propose a new algorithm named PWLS-TGV-MLP for simplicity that combines penalized weighted least-squares using total generalized variation (PWLS-TGV) and an artificial neural network model named multilayer perceptron (MLP) in this paper, we present this method briefly and evaluate it by a simulation experiment. The results of our simulation experiment on CT images indicate that the proposed algorithm efficiently recovers images and presents advantages that are slightly better than the traditional reconstruction approaches.