Free software - views on star reduction sought using Siril, Gimp & Starnet++

[Zummerzet_Leveller]

Hi all.  Being a relative newbie to astrophotography the biggest challenge I find in processing is star reduction.  I am yet to push the button on paid for dedicated astro-processing software (other than Topaz AI) so have been developing my processing in Starnet++, Siril and Gimp.

So far my process goes as follows:

In Siril - Background extraction, colour calibration, green noise removal, Asinh and histogram stretch.  (Repeat as required).  I then make a starless version in Starnett++, repeat some of the steps on this in Siril before shipping the star version and starless into GIMP.  In Gimp I make a star mask and a separate star layer for the bright stars, mask any nebula in the starless image and do a noise reduction then mask and sharpen the nebula with a high pass filter.  In the starry version I darken the shadows, run the value propagate filter for more black (trying not make too many strange artefacts by playing with the settings), do some noise reduction, all while the star mask is applied.  I then layer this on top of the starless and set the starred layer to addition (I find screen brightens too much).  I then play with opacity until I get a result I am happy with.  Create a new from visible and then make some final tweaks, I sometimes tweak in Siril too if I want some more background extraction or better colour calibration.

Has anyone else got any tips for better star reduction in free software or those mentioned above?  I'd like to improve this process as this doesn't give the smoothest of backgrounds, the value propagate filter can get a bit blobby.

Here's what I achieved with the Cygnus loop - Asi2600mc and Redcat51 - 5hrs of data.  Starry and Star reduction version.

[vlaiv]

I would say that best way to do it would be to create starless and stars only image.

You need software that does math operations on image for that - I recommend ImageJ as free option.

Take linear, color calibrated image out of Siril and open it in Gimp.

Do levels - but only move right slider to the left - this will make image brighter but will still keep it linear. Don't over do it as stars will start to bloat if you over do it. This will be your "base image".

You don't need to see nebulosity at this stage. We only do this because StarNet++ (last time I checked) can only work with 16bit data and we want it to give us starless version while still linear.

Save base image from Gimp and apply StarNet++ on it to get starless version.

Once you have Starless version - subtract it from base image to get "stars only" image . This is best done in software that can do image pixel math, but you can also try to do it in Gimp by loading base image and starless image as layer on top of it. Turn layer mode to subtract and hopefully that will do the trick as well to produce "stars only" image.

Now you have two images - starless and stars only, which you then process separately. Starless you can stretch more aggressively and denoise / sharpen, whatever you like, and stars only image - well, you might as well keep it as is or give it very slight stretch to show fainter stars as well. Maybe apply round of sharpening on it to reduce star size. It is important that you leave background black in Stars only image - as there is no background there.

Once you have two versions that you like - put stars back on top of starless version as layer (mode addition or which ever gives you best results).

Btw, less is more - don't over process your images. Out of the two - I like version with normal stars more as one with reduced stars looks too over processed.

[Zummerzet_Leveller]

Thanks for the great feedback and insight @vlaiv.  Totally agree that the 2nd image is over processed, too heavy on the sharpening etc. I do want to be able to see the nebulosity better it the first image as it seems lost to me in the stars, that may be due to star bloat and poor control due to my processing approach.  I'll look into ImageJ, are there any useful guides to using the software anywhere?

When I've tried the subtraction method in GIMP the results have been pretty bad with all sorts of strange star colours.  I'll try again with an only brightened stretch on the star layer to see if that improves the subtraction method and I'll set about learning the processes in ImageJ too.  Thanks again for your insight. 

[vlaiv]

3 minutes ago, Zummerzet_Leveller said:

I'll look into ImageJ, are there any useful guides to using the software anywhere?

Not that I'm aware of.

Its general purpose scientific image processing software - primarily made for microscope imaging rather than astronomical. I found it useful because it allows for many operations to performed on images (a lot of math stuff) and it also allows for scripting and has many plugins available (look up Fiji as well - it is the same thing / distribution of ImageJ but preloaded with plugins so you don't need to get them separately).

[Zummerzet_Leveller]

4 hours ago, vlaiv said:

Btw, less is more - don't over process your images. Out of the two - I like version with normal stars more as one with reduced stars looks too over processed.

How is this in comparison?  This is a slight variation on the process you suggest above.  I had to push the stretch a bit further as Starnet++ was giving me horrid artefacts.  Managed to carry out the subtraction in Gimp for the stars layer too.