Decon behaviour

[jimmy101]

Hi Ivo,

When using decon I always find that I cannot have a radius size greater than 1.7pix (when using an image scale of around 1.25"/pix) otherwise I get dark halos around small stars. The thing is the nebula responds well to a value of up to 2pix, but I cant push it because of star halo. Any tips?? I usually run 12 iterations.

The other odd thing is that I get a dark "X" from through small stars using decon unless settings are very mild (say 1.6pix radius). This behaviour can be mitigated by masking all stars. I find Deconv works far better if I:
Create a Fat Star mask - Grow x5
Create a star mask - add new mask to old - shrink x2
Grow blog any big stars that might need more coverage.

I don't get any "X" through stars as the core is protected using the mask, but the stars still tighten up around the edges. Any other way to do this? Also perhaps Decon should be protecting all stars?

Also I notice in PixInsight they recommend 20-30 iterations - Should I be doing this also? I'm guessing if I don't see much changing after 10 there is no point.

[admin]

Hi Ivo,

When using decon I always find that I cannot have a radius size greater than 1.7pix (when using an image scale of around 1.25"/pix) otherwise I get dark halos around small stars. The thing is the nebula responds well to a value of up to 2pix, but I cant push it because of star halo. Any tips?? I usually run 12 iterations.

The other odd thing is that I get a dark "X" from through small stars using decon unless settings are very mild (say 1.6pix radius). This behaviour can be mitigated by masking all stars. I find Deconv works far better if I:
Create a Fat Star mask - Grow x5
Create a star mask - add new mask to old - shrink x2
Grow blog any big stars that might need more coverage.

I don't get any "X" through stars as the core is protected using the mask, but the stars still tighten up around the edges. Any other way to do this? Also perhaps Decon should be protecting all stars?

Also I notice in PixInsight they recommend 20-30 iterations - Should I be doing this also? I'm guessing if I don't see much changing after 10 there is no point.

Hi,

I will send you a link to 1.5 RC6 in which this was addressed/re-worked via PM. Decon as a whole should also be a tad faster.
The amount of iterations really depends on the particulars of the decon implementation's regularization scheme. For one, PI implements a more naive decon implementation that does (cannot) not take local signal quality into account (e.g. it assumes blanket parameters) for the whole image. This means it cannot be as targeted/aggressive per iteration.

As a result you will find you only need a few iterations in StarTools for relatively low SNR scenes (e.g. DSO images). Planetary/lunar images may still benefit from more iterations, as there is more signal to work with and refine.

As of 1.5 you can use the post-tweak/pre-tweak parameters to evaluate any difference between the amount of iterations, so this may help with estimating effectiveness.

Do let me know if this solves (or alleviates) the problem you're seeing!

Thanks!