9 replies to this topic

[dividebysandwich]

I am having a problem in my image processing with Siril, and I'm sure it's user error. I was hoping someone could enlighten me.

I'm using a Touptek ATR2600C with a Stellarmate Pro and quite often I end up not being able to process the resulting data with starnet because the resulting starless image is way too overblown and bright. I use gain 100, offset 110, 120 second exposures, in this example about an hour of data.

 

This is what my stacked, background extracted and color corrected image looks like with autostretch:

 

 

 

If I use starnet with pre-stretch enabled, I get this resulting starless image:

 

 

I have also tried pre-stretching myself and didn't get very far with that.

 

My process is usually the built-in preprocessing script, I am using flats and bias frames (no darks), then after stacking I do plate solving, photometric color calibration, background extraction with hand-placed sample points avoiding bright stars and nebulae, then I remove green noise.

I tried using histogram and GHS on the starless image but I can't get anything usable out of it. I tried the linear BP adjustment in GHS to lower the black point but I either lose all detail or everything is washed out in gray.

 

 

My questions are:

 

1) Am I doing something wrong with my basic process?

2) If not, is the resulting starnet image something that you can work with? If so, how?

 

I have uploaded the following original files in case someone is willing / interested to take a look :

 

result.fits - stacked image

result_bge_rgn.fits - color calibrated, background extracted

starless_result_bge.fits - the output of starnet (prestretch was checked)

starmask_result_bge.fits

 

I also included the stacked flat and bias masters, and in the "lights" directory there's the original light frames, just in case I should be doing something different there.

 

 

 

[UKnewbie1729]

1) No you're not doing anything wrong, thats what my starless images come out looking like too.

 

2) Have a look round the imaging forums, there are a lot of very helpful threads about getting the best out of starnet, the one tip I recall is do the stretching yourself using GHS but in many small steps, ISTR the thread starter said they used 20 small stretches to go from the linear image to the final stretched version.

 

(I am very new to using starnet, I'm sure others on here will be able to give you much more specific help/advice.)

[whiskeyzebra11]

Beginner here so take this with a grain of salt, but that offset of 110 seems quite high. Any particular reason it is set like that. I use an Asi533 with an offset of 20 or 40.

Do you know how many pixels are getting saturated when you capture the image? In Nina, on the imaging tab you can take a sample image and see how many pixels are exceeding the max threshold. You can experiment with gain 0 at 2 minutes, or gain 100 at 60 seconds, just to see how that impacts the number of saturated pixels. Just something to check.

I typically use Siril with Starnet and have three approaches:
- in its linear state, remove the stars making sure to check the option to prestretch linear data

If that is bad, apply a histogram autostretch taking note in the console of the values applied. Then remove the stars (not checking the box to prestretch linear data) then run the -invmtf function (on the starless image) to invert the autostretch pasting in the values previously logged in the console.

Finally, if the above didn’t work well, back on the linear image, apply a light stretch to bring out the stars so they are visible, then try the separation, again not checking the box to prestretch linear data.

Usually one of these gives me the best result, or I may need to take the starless from one approach and the star mask from a different approach for the recombination.

I always use GHS to lower the blackpoint back to near the left edge of the histogram, but it’s usually already close to the left edge. I think that is due to my lower offset.

Edited by whiskeyzebra11, 19 March 2025 - 06:39 AM.

[prookyon]

I see that you have used Siril dev version. There is something messy going on there with color management.

What I usually do before starnet is to just remove the color profile completely and then run Starnet.

Tools menu -> Color Management -> Remove -> Close -> Starnet

Tried it with your file and the resulting starless image is probably more what you hoped for.

 

Edit: here is the result in linear mode

Edited by prookyon, 19 March 2025 - 06:49 AM.

[whiskeyzebra11]

Here is a version I did in Siril after a background extraction with Graxpert. Not sure if it is any better.

 

 

Then a very quick GHS stretch and recombination.

 

[dividebysandwich]

Thanks everyone for the quick responses, much appreciated.

 

 

I see that you have used Siril dev version. There is something messy going on there with color management.

What I usually do before starnet is to just remove the color profile completely and then run Starnet.

Tools menu -> Color Management -> Remove -> Close -> Starnet

Tried it with your file and the resulting starless image is probably more what you hoped for.

 

Edit: here is the result in linear mode

You are absolutely correct, this massively improved the starless result, thanks a lot for that hint! I still don't really know what I'm doing and following the various tutorials with varying levels of success, but not having to deal with an image that looks like I used the wrong washing machine setting helps a lot.

 

I'm in Bortle 4 and I assume that the lack of contrast in the nebula data is simply due to the short integration time, though I have to wonder, looking at the "1 hour challenge" images, whether I need to change image acquisition parameters.

 

Beginner here so take this with a grain of salt, but that offset of 110 seems quite high. Any particular reason it is set like that. I use an Asi533 with an offset of 20 or 40.

Do you know how many pixels are getting saturated when you capture the image? In Nina, on the imaging tab you can take a sample image and see how many pixels are exceeding the max threshold. You can experiment with gain 0 at 2 minutes, or gain 100 at 60 seconds, just to see how that impacts the number of saturated pixels. Just something to check.
 

An excellent suggestion. I think I was following someone else's advice for that camera, I should really try to establish my own value there. Having said that, are you sure that looking for the number of saturated pixels is the way to go? Isn't offset meant to avoid completely black pixels? Or did you mean saturated as in completely black?

[whiskeyzebra11]

 

An excellent suggestion. I think I was following someone else's advice for that camera, I should really try to establish my own value there. Having said that, are you sure that looking for the number of saturated pixels is the way to go? Isn't offset meant to avoid completely black pixels? Or did you mean saturated as in completely black?

Sorry, was thinking of two different topics:
1.) Is your offset too high? This is an interesting test to try if you happen to have Sharpcap (or similar program with a histogram). https://daleghent.co...g-camera-offset

2.) Regarding the star halos/artifacts, might they possibly be due to oversaturation? Just was wondering if oversaturation might make it harder for Starnet to deal with it cleanly. Not sure, just another thing to consider. I did run the Desaturation function in Siril and it found some saturated stars and attempted to fix them. The resulting separation was slightly better after that.

[danny1976]

Get StarXterminator and your starless images will be cleaner.

 

I always do a little bit of noise reduction before SXT but still in linear mode. The updated NoiseXterminator does wonders.

 

And 1 hour is not enough. Try to be patient and collect more data, 20h is a good starting point.

Edited by danny1976, 19 March 2025 - 03:59 PM.

[Drothgeb]

Get StarXterminator and your starless images will be cleaner.

 

I always do a little bit of noise reduction before SXT but still in linear mode. The updated NoiseXterminator does wonders.

 

And 1 hour is not enough. Try to be patient and collect more data, 20h is a good starting point.

I agree. I get much better results with StarX. Only it doesn’t work in Siril. It’s part of the reason I finally started using PI. But even it’s not perfect. Photoshop works best for correcting any left over artifacts.

[dividebysandwich]

Pixinsight + SXT is something I might have to invest in down the road. But for now I'm glad that clearing the color profile data made the starless image usable again.

Thanks a lot for the good suggestions, and don't worry, I know that one hour is not a lot of data to get great images out of. But hey, I'm learning and hopefully will get better with processing as I gather more data