Maximum-likelihood constrained regularized algorithms: an objective criterion for the determination of regularization parameters

Details Maximum-likelihood constrained regularized algorithms: an objective criterion for the determination of regularization parameters Maximum-likelihood constrained regularized algorithms: an objective criterion for the determination of regularization parameters

Dec 1999

adsabs.harvard.edu/cgi-bin/nph-data_query?bibcode=1999spie.3866..144l&link_type=abstract

Proc. SPIE Vol. 3866, p. 144-155, Optics in Atmospheric Propagation and Adaptive Systems III, Anton Kohnle; John D. Gonglewski;

Physics

Optics

2

Scientific paper

We propose regularized versions of Maximum Likelihood algorithms for Poisson process with non-negativity constraint. For such process, the best-known (non- regularized) algorithm is that of Richardson-Lucy, extensively used for astronomical applications. Regularization is necessary to prevent an amplification of the noise during the iterative reconstruction; this can be done either by limiting the iteration number or by introducing a penalty term. In this Communication, we focus our attention on the explicit regularization using Tikhonov (Identity and Laplacian operator) or entropy terms (Kullback-Leibler and Csiszar divergences). The algorithms are established from the Kuhn-Tucker first order optimality conditions for the minimization of the Lagrange function and from the method of successive substitutions. The algorithms may be written in a `product form'. Numerical illustrations are given for simulated images corrupted by photon noise. The effects of the regularization are shown in the Fourier plane. The tests we have made indicate that a noticeable improvement of the results may be obtained for some of these explicitly regularized algorithms. We also show that a comparison with a Wiener filter can give the optimal regularizing conditions (operator and strength).

Affiliated with

Also associated with

No associations

Rating

Maximum-likelihood constrained regularized algorithms: an objective criterion for the determination of regularization parameters does not yet have a rating. At this time, there are no reviews or comments for this scientific paper.

If you have personal experience with Maximum-likelihood constrained regularized algorithms: an objective criterion for the determination of regularization parameters, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Maximum-likelihood constrained regularized algorithms: an objective criterion for the determination of regularization parameters will most certainly appreciate the feedback.

Rate now

Comments { 0 }

Profile ID: LFWR-SCP-O-1062018