An Iterative Reconstruction For Tomosynthesis Imaging Using Non-Local Means

Abstract

This paper proposes a new 3D iterative reconstruction method for reducing out-of-focus slice blur in tomosynthesis imaging by combining two powerful denoising methods. The methods used in the reconstruction are based on Total Variation (TV) minimization and Non-Local Means (NLM) filtering. A new method (ART+TV) NLM is introduced by adapting both methods to Algebraic reconstruction technique (ART) which is one of the most commonly used iterative method in limited angle tomographic imaging. Simulations were performed on a designed 3D phantom. The method is shown to be more effective by keeping SNR higher and RMSE values lower compared to the ART and ART+TV reconstructions.