Pls Help:NGC6823 with Nebula, Canon DSLR, >12hours exp time, low contrast nebula

[Christoph Puetz]

Dear PI Community,

I took 179x5min frames of NGC 6823 and its surrounding nebula with a Canon DSLR
and a Celestron 8 telescope, stopped at f/6.3. Exposure time is over 12 hours.

Calibration and integration worked fine and gave me a SNR of about 12:1.
I would like to emphasize the surrounding nebula, wihich still has a very low contrast.

I tried several times and did not get a better result than this picture:
http://www.ccdsky.eu/Astronomy/Nebula/slides/ngc6823_sh86_v3.html

My problems are:
- if I try to enhance the nebula (i.e. CLAHE), I increase noise.
- in strict linear processing, I do not succeed to get a good result (or I have to experiment in the parameter range of 0.000001)
- I also tried wavelet processing and some masking, but get a lot of problems with deringing with the stars (is it possible to "remove" them temporarily ?)
- If I separate the channels into L/R/G/B and enhance the red- and luminance-channel, the re-combination increased noise.
- Noise reduction with ACDNR gave me some large-scale artefacts ...

So for me this is quite a challenge !
Therefore you are invited to fully process this image and - please - help me and my PI friends to share your
processing steps. I will also provide this image as a base for any tutorial, if needed.

You can download the linear FITS file here (link is valid for 14 days and 14 downloads, can be increased as needed).:
https://www.hidrive.strato.com/lnk/nOsE3gCV

Thanks in advance for sharing your processing hints with us  8)

Christoph

[Christoph Puetz]

OK, I will begin with a processing way that I found to be useful.

Step1: Split Integration Image in L-R-G-B Channels.
Step2: Most Nebula Information is on the red channel: With ATW I divided it in 2 Parts, a "low scale" part (layers 1-4) and a "high scale part" (layers 5 to residual).
Step3: Build a luminance mask for each scaled part with "range selection"
Step4: Enhance contrast with LHE on the masked parts.
Step5: Pixelmath: Scaled image1 (low scale) added with Scaled image2 (high scale), try some "weigthing factors" (i.e. scale1*2+scale2*3),
also interesting is to multiply both scales directly.
Step6: DarkStructure Enhance on Luminance channel
Step7: LRGB Combination
Step8: Optional : Clone the resulting RGB Image. Now use ACDNR and experiment with the "Dark and Bright Sides Edge Protection, esp. the "Overdrive Option". This can "wash some stars out and enhance the faint nebula structure". Then multiply the result with the image of step 7.
Step9: Use curves function for color correction.
Step 10: Maybe you can try to use the "GradientHDRCompression" for some enhancement of the fine nebula structure.

That's that I could get out so far. Might be a good result for such a small C8 and Canon DSLR.

[Alejandro Tombolini]

Hi Christoph,

I think the Canon has taken to its limits with this picture!

First I have used HistogramEqualization to see the information in the image and select a place where to make the preview for use in BackgroundNeutralization.



After that CBR to remove the horizontal banding in the bottom of the image



Once again helped with HistogramEqualization detect the places where to place samples for use with DBE. (Generate a clone, make samples over it, generate the DBE instance and then apply on the image)



ColorCalibration using the same preview used for BackgroundNeutralization as Background Reference image and the complete image for White reference image.



Generate the PSF and use it in Deconvolution protecting stars cores with starmask



A sequence of Masked Stretch until Target median of 0,4. This is time consuming when you save the project and I must apologize at this point because I could not save the project every step, and I can not remember if some other clips were applied in between in addition to the last shadows









With HT clip shadows



CurvesTransformation to increase contrast, saturations and colors



More contrast and Chrominace Noise Reduction using ATWT



SCNR to eliminate greenish



Noise reduction with L mask protecting stars using GREYCstoration



Still looking for more contrast use LHE with the same L mask protecting stars (At this time I detect some problems in stars cores, maybe the previous L mask was not good enough and was caused for GREYCstration



Now I will built a mask without stars to protect bright nebulae and apply curves. As R channel has the information, extract it and duplicate, applying ATWT on duplicate to remove layer bigger than scale of 256 pixel



PixelMath to substract R_clone from R unchecking "rescalete" and after that apply ATWT to smoothness and built the rangemask over it



Protecting bright nebulae apply CurvesTransformation to desaturate background



Now invert the mask and apply CurvesTransformation to saturate nebula protecting background



ACDNR to reduce noise



With the R channel protecting background apply LogStrech to increase nebulae



As the halos in stars has increase considerable i will tray to reduce and adjust colors in the stars. For that build a mask to protect background and stars cores but not halos
The mask is the result of BlendScript multiplying the starmask and rangmask



Apply CurvesTransformation in stars halos protecting with BlendMask



Final Unsharp without mask



Final Image:



Regards, Alejandro

[FunTomas]

Fantastic result Alejandro. I admire your result.

[Raymond]

Hi Christoph; thanks for providing this interesting image.

Alejandro - your processing and result are impressive. I appreciate these complete processing examples - they help me a lot in understanding PI and in improving my processing skills. After a first look through your processing, I have some questions:

1. MaskedStretch: You used target medians of 0.10, ..., 0.40. For what reason did you use these values? Can they be used in general?
2. Why didn't you use a noise reduction procedure before your first CT as this risks to increase your noise in the background too?
3. When using CT, you did a lot of changes in the different channels and components. After what criteria do you decide for all those changes? I have to admit, that I completely lost the overview 

Your routine in controling the halos is excellent too! I looked for quite some time for such a routine.
Thanks again ....

[Alejandro Tombolini]

Thank you FunTomas and Raymond!

they help me a lot in understanding PI and in improving my processing skills

  Excellent!

1. MaskedStretch: You used target medians of 0.10, ..., 0.40. For what reason did you use these values? Can they be used in general?

Usually when use MaskedStretch select 0,2 as Target median and 200 histogram transformations. In this particular image, as the nebula is too faint, I started trying whit a lower number for target median, as I realized that it was not enough, i have  continued increasing the target until 0,4, always with 200 histogram transformations to have very little changes in each iteration.   

Do not consider the values as general values, it will depend on the image. However if you use always the 500D and take sub of similar times, maybe will find a target median that work well as a first stretching.

2. Why didn't you use a noise reduction procedure before your first CT as this risks to increase your noise in the background too?

Whit my camera (Canon 450) I usually make a noise reduction with ATWT before stretching and works well, but in this case I did not want to risk losing with a noise reduction some detail that had not been seen. Christoph mentioned that had integrated 179 images and reached a good result in terms of SNR.
Note that the first Curves Transformation increase RGB but not in shadows, and the noise added in chrominance with "c" and "S" is possible to be removed with ACDNR or ATWT when the image is stretched.

3. When using CT, you did a lot of changes in the different channels and components. After what criteria do you decide for all those changes? I have to admit, that I completely lost the overview 

It was not a lot  Excluding Curves Transformation used to controlling the halos, note that I only use Curves Transformation to saturate or desaturate (c and S), RGB to increase light in medium and high levels and Hue to turn pinks in red.

Saludos, Alejandro.

[Christoph Puetz]

Hello Alejandro,

thank you very much for your kind investigation and your very detailed processing steps.
I am very astonished how you enhanced the faint structures.
Esp your halo-reduction and some hints to the "masked stretching" were very informative.

Once again  - I learned a lot more working in PI.
And it is great that there is so much support in this forum.  8)

In fact, this was an "extreme" photograph in combination with a DSLR and a quite small telescope.
Indeed - this was intended to experience the limits of DSLR techniques. Maybe this will lead to a cooled CCD camera in
the future.
My next try will be stephans quintett with a DSLR - but with a PLANEWAVE 50cm scope. If you find it interesting, I will
publish it here again.

Kind regards and thanks for your help.
Christoph

[Alejandro Tombolini]

My next try will be stephans quintett with a DSLR - but with a PLANEWAVE 50cm scope. If you find it interesting, I will
publish it here again.

WOW, of course Christoph, It sound very interesting! If you could achieve a good SNR image like this will be and excellent challenge.

Regards, Alejandro.

[cs_pixinsight]

Bravo Alejandro!  This is another excellent processing tour de force.  I love reading your processing tutorials, since they are always an education in processing within PI.

Craig

[Raymond]

Hi Alejandro
Thanks a lot for your explanations. I tried parts of your processing procedure on an image of NGC7331 - and that result is much better than my first attempt without your workflow  . And there is even the Stephans Quintett    !
WO 110/770 with Canon 500d (mod); total of 3h 50min; ISO 800; limit sky mag: 5.2.

[Alejandro Tombolini]

Excellent Raymond, congrats for the picture!  and good to know the example is usefull

Regards,
Alejandro.

[astropixel]

Delayed reply - Perhaps a sub-forum, entitled, very appropriately, "This is How You Do It: Astronomical Image Processing with Alejandro Trombolini," would be a good idea.

Beautiful image Christoph.

[Christoph Puetz]

Thank you all, guys. A sub-forum would be great, as Alejandro is a really great supporter for our questions.
PS: Stephans quintett (with 50cm Planewave) will surely come, but currently bad weather in Germany