Pixinsight and starnet++

[barrie greenwood]

Hi guys 

So when using starnet to reduce stars do I process the image to finished state before using starnet or at some point along the way

At What point do i remove the stars in the process .

Im using starnet as a Windows stand alone as I am unable to get it running in pixinsight

Regards Barrie 

[teoria_del_big_bang]

I think the idea is to separate he stars and nebula so you can stretch the nebula without over stretching the stars so they are saturated  and take over the image detracting from the nebula,  Also to do some colour saturation on the nebula without affecting the star colours without the need to use masks which are not always easy to create for all stars. So with that in mind it is done early on in the post processing and I usually do this in the linear stages.

Not sure why you cannot get it working in PI though. Are you using Starnet2 ?  Maybe explain what the issue is, that you cannot download it so it appears in the process menues or that when you use it it fails in some way ?
Also maybe try StarXTerminator, I am finding it works better than Starnet2.

Steve  

[barrie greenwood]

So after I stretch it I split the photo into two one with stars and one with out then work on the starless one and recombine them both is that the idea ?

[teoria_del_big_bang]

From what I read people seem to do it at slightly different stages, but the consensus seems to be very early on in the post processing, either just after the RGB combination, or maybe after a moderate stretch so the stars are visible but not overblown.
Personally and doing it in PI I do it in the linear stage, shortly after RGB combination, having done some initial noise reduction and DBE before the RGB combination but before any stretching.

You can do some searching on Google but probably there is no definitive answer to say you must do it at this precise stage, and I have done a bit of experimenting and so long as you do it early on I am not even sure it makes a difference.
I am no expert on any of this and have searched as well to see when and how to do all this.

I certainly would persevere with getting it (0r StarXTerminator) working in PI it would be so much easier.

But to answer your last question

11 minutes ago, barrie greenwood said:

So after I stretch it I split the photo into two one with stars and one with out then work on the starless one and recombine them both is that the idea ?

The answer is basically yes but for me remove before the stretch - but that is not what everyone does.

Steve

[barrie greenwood]

Ok slowly getting it but I use a dslr so it's a one shot colour affair .

Or is my workglow wrong as  I don't split the channels I've looked online but there is so much I don't really understand 

[Budgie1]

I also use a OSC camera and I either run the StarNet process right before I stretch the image, or right after.

I remove the stars before the stretch to allow me to stretch the stars less than the background, especially if there are a lot of stars or some really bright ones.

If there are few stars or they're all basically the same brightness, then I remove them after the stretch. As Steve has said, there is no right or wrong time to use the process, just see what works for you.

There's a StarNet2 installation & usage tutorial HERE which may help if you're still having issues getting it working in PixInsight.

Edited October 15, 2022 by Budgie1
Spelling

[barrie greenwood]

Thanks guys I'll give it a go in the morning I've had a play about with it and ended up with this 

Its a long way from great but my flow isn't great either .

So I did the bg removal then background neutralise then photometric colour calibration then stretched it .

Took the results into starnet and split the stars then into photoshop cs6 and combined the two in layers and allowed the star version to come through by adjusting the opacity of the starless version .

Am I on rhe right track or wayyyy off 🤔 

Again thanks for the time taken to reply gys 

Edited October 15, 2022 by barrie greenwood
Spelling

[Budgie1]

I think you're on the right track but, if you can get StarNet2 working with PI then your life will be easier.

In PI, I use PixelMath to recombine the background & star images into one. I use one of two formulas, either

backroundimagename+starimagename

or

~(~backgroundimagename*~starimagename)

The second one does tend to give the better results and is what I use most of the time now.

[barrie greenwood]

I got up early today and managed to integrate starnet into pixinsight and it does work however the second formula gave me a weird result that was white like a starmask

Regards 

[ollypenrice]

9 hours ago, barrie greenwood said:

 

Am I on rhe right track or wayyyy off 🤔 

 

I don't know about 'way off,' but 'off,' yes. (Sorry! ) What we see in your image is not what you were aiming for, I don't think. You have quite large stars applied at what looks like partial opacity. What you want, I imagine, is the reverse, small stars of full opacity and brightness.

I use StarXterminator in Photoshop rather than Starnet++ in Pixinsight but the underlying principles must be the same.

If working on an existing processed image I remove the stars and then follow the method explained by Xiga in the thread linked below.  When I have the original stars as a layer on top I have to reverse-process them to make them smaller. Basically this just means using Levels and moving the grey point slider to the right to de-stretch them. It can also help to give them a mild Gaussian blur to integrate them into the image more naturally. Although de-starring a finished image probably isn't the best way, I find it works astonishingly well.

If starting from scratch I've had good results by stretching the image gently until I feel the stars are as I would want them at the end. At this point I de-star and save the stars as they are. I then continue to stretch the starless and can push it considerably harder than I'd be able to do with the stars in place. As Xiga suggests, I put them back as a top layer over the starless in Blend Mode Screen. So I guess what you're looking for is the Pixelmath equivalent of Blend Mode Screen. This will certainly be available in PI and is probably what Budgie 1 is suggesting.

I'd certainly persevere, though, because this is the biggest breakthrough in astrophotography that I've seen since I began thirteen years ago - and by a country mile.

Olly

 

Edited October 16, 2022 by ollypenrice
Clarification

[teoria_del_big_bang]

1 hour ago, ollypenrice said:

f starting from scratch I've had good results by stretching the image gently until I feel the stars are as I would want them at the end. At this point I de-star and save the stars as they are. I then continue to stretch the starless and can push it considerably harder than I'd be able to do with the stars in place. As Xiga suggests, I put them back as a top layer over the starless in Blend Mode Screen. So I guess what you're looking for is the Pixelmath equivalent of Blend Mode Screen. This will certainly be available in PI and is probably what Budgie 1 is suggesting.

Do you think this initial stretch to get the stars correct will yield a different result to removing before the stretch and then individually stretching each image before some form of blend ?
Although thinking about it now 🤔, maybe this is a better workflow if only for the fact you are stretching the image with the stars included and so get a better sense of what the final image is with regard to star size and brightness as you have the whole image to view in real time, where as an individual star stretch is just against a blank background., making it difficult to judge when you have it just right.
So with that in mind I think I will adopt that kind of workflow.

I too am still trying to in down a good workflow that works for me so just a couple of questions if that#s okay 🙂 

• If you had to do some work on the stars then, such as NB halo removal, or some form of sharpening would you do that on the stars only image (or could that yield some artefacts as the star size will change significantly between removal and the blend) or do that on the whole image before removal ?
• And I know you are very careful not to do too much regarding noise removal but if you do use it where in the workflow would you do this ?

Steve

 

Edited October 16, 2022 by teoria_del_big_bang

[teoria_del_big_bang]

2 hours ago, barrie greenwood said:

I got up early today and managed to integrate starnet into pixinsight and it does work however the second formula gave me a weird result that was white like a starmask

Regards 

Is that starnet2 ? 
I would use Starnet2 not Starnet. EDIT: Ah just read the thread header again and you refer to Starnet++ so that's the same as Starnet2 so sorry wont bang on about that anymore 🙂 
Not sure why the error, only time I can recall strange results is if I try to use the process on a linear image and not tick the checkbox to say linear image, or vica versa.

Also I know there is a small cost to buy StarXTerminator, but it is not much and I have found this works best for most images I have tried so far compared to Starnet2. I think there is a trial period for StarXT so I would recommend at least giving it a go.
Maybe Starnet2 will work better on some images, not sure but it is good to have both processes so you can at least try both and see if one yields better results than the other, it is worth the extra bit of processing time.

Steve

Edited October 16, 2022 by teoria_del_big_bang

[ollypenrice]

1 hour ago, teoria_del_big_bang said:

Do you think this initial stretch to get the stars correct will yield a different result to removing before the stretch and then individually stretching each image before some form of blend ?
Although thinking about it now 🤔, maybe this is a better workflow if only for the fact you are stretching the image with the stars included and so get a better sense of what the final image is with regard to star size and brightness as you have the whole image to view in real time, where as an individual star stretch is just against a blank background., making it difficult to judge when you have it just right.
So with that in mind I think I will adopt that kind of workflow.

I too am still trying to in down a good workflow that works for me so just a couple of questions if that#s okay 🙂 

• If you had to do some work on the stars then, such as NB halo removal, or some form of sharpening would you do that on the stars only image (or could that yield some artefacts as the star size will change significantly between removal and the blend) or do that on the whole image before removal ?
• And I know you are very careful not to do too much regarding noise removal but if you do use it where in the workflow would you do this ?

Steve

 

I don't know whether a linear image or a soft stretch is best before star removal, not having done enough testing. However, the thoughts you list after this question are the same as my own. I do like to have eyes on what I'm doing, which is why I prefer Photoshop.

I never sharpen stars. I only ever soften them - and especially when recombining them after removal. I tend to think they look too hard when seen as a top layer prior to recombining. I'll certainly sharpen nebulosity on the starless, though, to bring out structure. I'll also make heavy use of local contrast enhancement. This might be LHE in Pixinsight, Noel's Actions or stretching through a Layer Mask made from an equalized, blurred copy of the image. This may be an inverted mask or not, depending on where I want the extra stretch.

If a de-starred image has residual evidence of halos (as sometimes happens) I make no bones about using the spot healing brush to get rid of them.  I think you just need to check whether the halo is coming from the star layer or the starless before thinking out an approach.

I use Noise Xterminator on a  boottom copy layer of the starless image before going for the hardest later stretches. I'll then use Colour Select to pick up the noisy parts and erase them. I'm impressed by the lack of damage to detail in Star Xterminator but I'm still fussy about only applying it to parts which need it.  As you say, I really don't like too much NR. I'll even add nose to small parts of an image which have become too smooth in processing.

I've just had another great result (well, I think it's great! ) by de-starring and re-starring a fully processed image, reverse-processing the stars once separated. (ie de-stretching them) This makes me think the process is very tolerant of when we remove the stars.

One thing to watch, though: if you de-star a mosaic in which the background isn't perfectly seamless, StarX will make it worse. The best way to check mosaic backgrounds is to equalize the image for a check

Olly

[teoria_del_big_bang]

Thanks @ollypenrice that gives me some great advice and helps a lot 🙂 

Steve