A projected Newton-CG method for nonnegative astronomical image deblurring (Q937181)

The authors study an optimization problem arising in the process used for deblurring astronomical images. In the first section they present the necessary background to this interesting problem including the extent of corruption of astronomical images, the image reconstruction process and the associated convex minimization problem that needs to be solved -- which is ill-posed and the objective function needs to be regularized. In the second section the main method for solving this problem is described which is a projected Newton conjugate gradient (CG) approach. Several properties of the method are presented, with proof, including theorems on the convergence of the algorithm. The article concludes with a section of computational results where simulated blurred astronomical images are reconstructed using various values for the parameters of the algorithm.