Back to forums

Getting my nebulas to shine

Processing Request for feedback Sh2-174 Planetary Nebula
17 replies•713 views
Torben van Hees avatar
Torben van Hees avatar
Hi,

I have a very specific question: I have recently imaged two of the larger planetary nebulas.

Sh2-174
and

PK 164 +31.1
While I do like the overall brooding and dark outcome, I would think that it should be possible to give them a bit more brightness. This is a problem I feel I have with several of my images. Of course, this could be due to too little integration time, which is why I chose these two rather long exposures as examples. Any good tips to brighten these images up? When I just stretch more, the background noise starts to show up and my curves adjustments tend to produce posterization effects. I feel like I am missing something.
Well Written Engaging
John Hayes avatar
John Hayes avatar
Those look like very nice images.  The way that you display them is totally dependent on how you process the data.  You didn’t say what processing programs you use but each program provides its own tools for stretching the data, enhancing detail, and controlling noise.  You clearly know how to process the data enough to get to the results you have but you may be missing some of the finer points on how to control brightness through the stretch.  Perhaps the best advice that I can think of would be to check out some processing videos online.  There are a lot of them and a lot of folks provide their own processing steps in gory detail.

The one thing that you said that sounds worrisome is that you see posterization effects when you stretch more aggressively.  How are you stacking?   These images clearly have enough exposure time (and subs) that you shouldn’t be running into quantization issues.  Your stacked data should easily achieve 32 bits so you shouldn’t see posterization effects unless there is something that’s limiting the output.  The NB signal for these two objects is indeed very low, which makes dealing with noise a priority, but I don’t believe that you should be running into posterization effects.

John
Helpful Insightful Respectful Supportive
Torben van Hees avatar
Torben van Hees avatar
Thank you for your input, John. It may well be that I am missing the finer points of stretching. I will follow your advice and seek out a video tutorial on that part of processing specifically. Also, I have now checked out some more images on SH2-174 and it seems that the really good ones have 2-4 times as much integration time as mine - way to go, too.

I am using PixInsight for almost all my processing as its more maths-based workflow suits my way of thinking far better than something like Photoshop. I am not good at getting post processing results from APP, although I really like its multi-session calibration routines.

Sh2-174 was calibrated and stacked with WBPP, while I did manual calibration and image weighting on PK 164 +31.1 (because the camera was new to me). Linear workflow was short: MureDenoise, Crop, Deconvolution. Then ArcSinHStretch (for Sh2-174)/MaskedStretch (for PK 164+31.1), ACDNR and MLT denoise (both heavily masked), a small amount of more stretch and black point adjustment with HistogramTransformation, then CurvesTransformation on the L (stars masked). It basically is this final step that I am struggling with: I feel like I need to adjust the curve to a steep gradient between the darker and lighter parts to brighten up the image, but that leads to posterisation. Maybe something in my processing before destroys the gradient there. I tried to leave out Deconvolution and replaced MureDenoise with other denoise algorithms, just to check if these are not responsible, but that did not change the outcome.
Helpful Respectful
John Hayes avatar
John Hayes avatar
Ah…you said something that rang a bell.  ArcSinHStretch is a fantastic tool for controlling star color but in my experience, it tends to produce a very dark result for the subject object.  I’ve found that the same goes for MaskedStretch.  You might try some experiments setting up the ScreenTransferFunction to show what you want then to drop those settings onto the HistogramTransformaion.  I think that you’ll find that the object will appear much brighter but that the stars won’t be as well controlled.  There are ways to fix that so this is just the first step to get the object stretched more closely to where you want it.  If you can get that to work then you have to figure out how to handle the stars—and that’s something that everyone has to struggle with for a while before getting it dialed in.

John
Helpful Insightful Respectful Engaging Supportive
Victor Van Puyenbroeck avatar
Victor Van Puyenbroeck avatar
While arcsinh is a very powerful tool to compress dynamic range, it also reduces contrast. What works well for me, is to duplicate the linear image after all linear processing steps. Then stretch one version with HistogramTransformation for strong brightness and contrast. The other copy is stretched with Arcsinh or Autohistogram with logarithmic stretch, because it produces very clean stars on my Newtonian. Finally, the stars are copied from the Ash version to the HT version using a starmask.

I also noted that the background in your images above is almost completely black. That might explain why you need to make such steep curve adjustments.
Well Written Helpful Insightful Engaging Supportive
Torben van Hees avatar
Torben van Hees avatar
Thank you, John, I will try that. And yes, it seems I started to get these darker results (but better stars) with these two tools. Stars are an ongoing issue, and from what I read, I believe they will remain one, but I am not confident enough there yet to ask the right questions.
Torben van Hees avatar
Torben van Hees avatar
Victor Van Puyenbroeck:
While arcsinh is a very powerful tool to compress dynamic range, it also reduces contrast. What works well for me, is to duplicate the linear image after all linear processing steps. Then stretch one version with HistogramTransformation for strong brightness and contrast. The other copy is stretched with Arcsinh or Autohistogram with logarithmic stretch, because it produces very clean stars on my Newtonian. Finally, the stars are copied from the Ash version to the HT version using a starmask.

I also noted that the background in your images above is almost completely black. That might explain why you need to make such steep curve adjustments.

I will try that approach to combine ArcsinH with HT, thanks for the tip. As for the background: It tends to have some colored cast even after background neutralisation. Maybe I should apply desaturation to the background and lift it up some.
Well Written
HR_Maurer avatar
HR_Maurer avatar
Hi Torben,
nice images! If i had longer focal distance, i'd have them right on top of my list. Maybe i'll try if i can get my Dobson into astro photography…

I think it might not necessarily be the contrast in the nebulae, which misses the last piece of punch. I might be wrong, but if you bring the stars to a more shining appereance, the whole composition could benefit. There has already been some discussion about star stretching strategies. It's worth a try!

CS Hoschie
Respectful Supportive
Torben van Hees avatar
Torben van Hees avatar
Thank you Mike and Hoschie, I am just now trying a Lum extraction after channel combination (which I forgot to mention in the workflow). And I agree some more shiny stars would improve the overall appearance - stars are difficult, though.
Torben van Hees avatar
Torben van Hees avatar
Thanks to everyone for helping. I have reprocessed with what I believe is the best version that I can produce currently. I am not totally happy yet with the bigger stars, but I feel that the nebula is much improved.

Sh2-174
Well Written Respectful
Kevin Morefield avatar
Kevin Morefield avatar
John Hayes:
ArcSinHStretch is a fantastic tool for controlling star color but in my experience, it tends to produce a very dark result for the subject object.  I’ve found that the same goes for MaskedStretch.


I second this statement.  I use those two stretches for color only and use an STF/Histrogram stretch for the Luminance component.
Thomas avatar
Thomas avatar
I've recently been using the following steps with some decent success: before stretching I duplicate the image and do the screen transformation/histogramtransformation stretching mentioned above on the duplicate copy (intent here is to make some masks). Take that image and use Starnet in PI to split it into a star mask and then a starless version. I'll then do an aggressive stretch on the starless version to really bring out all of the nebulosity and a slight stretch on the star mask to brighten up the stars. Now with my unstretched nonlinear image, ill do a maskedstretch on that one to get it nonlinear ( I like the look of my stars better with the maskedstretch). From here I'll use the starless and starmask versions created from the duplicate copy earlier on to apply as initial masks for stretching.
Helpful
Rich Sky avatar
Rich Sky avatar
I am not an expert, I prefer your first image (original post) to the second(image in a later post) The second image appear on my PC oversaturated.

Clear skies
DarnitsCloudy avatar
DarnitsCloudy avatar
Are you sure your NB combination is not causing stars to go off color?
The method described here: How to combine Ha and Color in PI (Bubble Nebula) by The Elf, helped me keep natural star colors when adding Ha data.

/Cloudy
Torben van Hees avatar
Torben van Hees avatar
Are you sure your NB combination is not causing stars to go off color?
The method described here: How to combine Ha and Color in PI (Bubble Nebula) by The Elf, helped me keep natural star colors when adding Ha data.

/Cloudy

What happened was that Starnet missed some of the smaller stars which turned to the nebula colours. I do not mind that much, as I am currently not so worried about the stars. If I wanted to, I could remove the small ones from the mask manually. I also did not saturate the stars as much turning them a more uniform whiteish color. The Elf's process is not applicable here at all since the RGB stack did not contain any signal from the nebula, it is much too faint for that.
Helpful Insightful Respectful
Torben van Hees avatar
Torben van Hees avatar
Rich Sky:
I am not an expert, I prefer your first image (original post) to the second(image in a later post) The second image appear on my PC oversaturated.

Clear skies

Yes, I think I overstretched some of the brighter parts. The more tuned-down version is now marked as final. Still struggling to reveal the fainter parts of the nebula well.
Well Written
DarnitsCloudy avatar
DarnitsCloudy avatar
Torben van Hees:
Are you sure your NB combination is not causing stars to go off color?
The method described here: How to combine Ha and Color in PI (Bubble Nebula) by The Elf, helped me keep natural star colors when adding Ha data.

/Cloudy

What happened was that Starnet missed some of the smaller stars which turned to the nebula colours. I do not mind that much, as I am currently not so worried about the stars. If I wanted to, I could remove the small ones from the mask manually. I also did not saturate the stars as much turning them a more uniform whiteish color. The Elf's process is not applicable here at all since the RGB stack did not contain any signal from the nebula, it is much too faint for that.

Ah, I see - Good that you found it. 
That said, the process I linked to does not require the RGB process to contain any nebulosity at all.  I have used it successfully to combine OSC stacks that had no visible nebulosity, with Ha stacks.  You just need them aligned of course.
Helpful Concise