Reconstruction of Images with Poisson Noise

This paper addresses quantitatively the problem of influence of statistical uncertainties embedded in the recorded image on uncertainties of the reconstructed image. In the analysis we use iterative maximum likelihood algorithm Andril (described by Sylwester and Sylwester 1998) developed for massive deconvolution of flare images obtained by the Soft X-ray Telescope (SXT) on Yohkoh. We illustrate the "ill-conditioned" nature of the image reconstruction problem and suggest the ways to reduce, at least partly, propagation of noise to the reconstructed image.