Many deconvolution cycles

[Ignacio]

Hi,

Is there any merit in applying more than one deconvolution process in a row? That is, do the whole process of deringing support, sampled PSF, etc., 2 or 3 times in a row, with say, 20 iterations each, instead of one with 60 iterations.

I've seen that some people do this (Ken Crawford, for example, using CCDSoft deconv.)

Just wondering...
Ignacio

[Juan Conejero]

Hi Ignacio,

If you are using deconvolution with physical justification, then applying it more than once makes no sense at all. Deconvolution is like solving an equation: you can only solve it once. If you manage to model the PSF of the image correctly, then deconvolution will do what it has to do with the linear data, not more (and not less!).

If you are using deconvolution as a cosmetic tool, then everything is possible, of course. Nothing prevents you from using deconvolution as a super-fancy sharpening filter, but there are many other algorithms and tools specifically designed for local contrast enhancement that can do a much better job with much less effort.

[Ignacio]

Thank you, Juan.

On a linear image, I understand your point about solving a mathematical equation. But when implementing a numerical/iterative approximation, things are less obvious. Otherwise, there would be no guessing/trial-and-error when setting parameters such as "number of iterations". The question was in that spirit, given that respected astroimagers do this on linear masters.
best
Ignacio