Limited-angle Tomographic Imaging Using a Constrained Natural-pixel Decomposition

Abstract
We present a method for tomographic image reconstruction, when the angle of view of the projections is limited. We use a decomposition into ‘constrained natural pixels’, covering the beam paths, weighted by a function that can incorporate various a priori information available about the object. The linear system resulting from this decomposition is inverted by a fast-converging iterative method. Numerical results show the importance of the added constraints, and prove the stability of the method in the presence of noisy data.