Hi, how you know it is the optimal solution ? What is your stop criterion. 

I figured it out in the meantime, I was using different boundary conditions.

Thanks anyway!

Dear Martin Gal,

Look the link. May be useful.

http://www.mathworks.com/help/images/examples/deblurring-images-using-the-lucy-richardson-algorithm.html

Regards, Shafagat

Hi, these papers may be useful:

http://ao4elt3.sciencesconf.org/13505/document

ftp://ftp-sop.inria.fr/pub/rapports/RR-5272.pdf

http://ceur-ws.org/Vol-446/p400.pdf

Since I am one of the authors of the cited paper (I suppose that on  boundary effect correction) I would like to understand the question.

RL is a slow algorithm and the limit is an image consisting of bright spots over a black background (good for restoring a star cluster, for instance); in general one needs early stopping (no generally accepted stopping rule exists) and a "visual" stopping is the best criterion.

If you need only 10-25 iterations I should say that your image is very noisy because the number of iteration increases if the noise decreases.

I am surprised that the matlab code requires more iterations because, if the code has not been changed, it is an implementation of the acceleration of RL provided by Biggs & Andrews.

Anyway, if you want a discussion of RL, its acceleration by SGP with and without boundary effect corrections, the IDL codes etc. you can look at our paper

M. Prato, R. Cavicchioli, L. Zanni, P. Boccacci, and M. Bertero, 2012, Efficient deconvolution methods for astronomical imaging: algorithms and IDL-GPU codes, Astron. Astrophys., 539, A133

You can download it from my web page

http://www.disi.unige.it/person/BerteroM/publications.html