Reconstruction from Limited Angular Projection Data Using Constrained Optimization

In this paper, a new type of pixel decomposition is introduced to suppress modeling noise and then an algebraic method is developed for reconstruction from limited angular projection data. The basic idea with this method is to directly control the tradeoff between resolution and noise, taking into account the uncertainty of the discrepancy criterion generally used in an algebraic method for reconstruction. The results of a simulation using a simple mathematical phantom are also presented, demonstrating the effectiveness of this method.