TheSkySearchers

Welcome to:

AP Processing Challenge No. 9

For this challenge Bradley Craig has submitted his narrow band, widefield images of the North America and Pelican nebulas.

From Wikipedia:

On October 24, 1786, William Herschel observing from Slough, England, noted a “faint milky nebulosity scattered over this space, in some places pretty bright.” The most prominent region was catalogued by his son John Herschel on August 21, 1829. It was listed in the New General Catalogue as NGC 7000, where it is described as a "faint, most extremely large, diffuse nebulosity.” In 1890, the pioneering German astrophotographer Max Wolf noticed this nebula’s characteristic shape on a long-exposure photograph, and dubbed it the North America Nebula. In his study of nebulae on the Palomar Sky Survey plates in 1959, American astronomer Stewart Sharpless realised that the North America Nebula is part of the same interstellar cloud of ionized hydrogen (H II region) as the Pelican Nebula, separated by a dark band of dust, and listed the two nebulae together in his second list of 313 bright nebulae as Sh2-117.

https://drive.google.com/drive/folders/ ... sp=sharing

Just click the link above to download and process Bradley's image using your preferred software and methods. There are three images, H, O and S. You can map these Hydrogen, Oxygen and Sulphur to any RGB channels of your choice.

AP Processing Challenge No. 9 starts today and ends on the 31st October.

Please keep the submissions coming in for future challenges. Just send a PM to a Mod or an email to tssapchallenge@theskysearchers.com

And don't forget Challenge No 8 is still running. If you haven't already, have a go at processing John Lowjiber's excellent Solar image.

Have fun!

Regards

Graeme

Thanks Graeme!

So for the narrowband inexperienced, the combined image, which is Sii to the Red channel, Ha to the Green channel and Oiii to the Blue channel in some fashion. I did my original with a straight 100% Sii to R, Ha to G and Oiii to B. There are some other combinations that may be pleasing as well.

Generally, this is a pretty easy one to process as there is a good bit of signal across all 3 narrowband channels. For the newbies, you will see an overly green color image when first combined, due to the large dominance of Ha which is applied to the G channel. Next will be an unwanted Magenta hue. These are the two main color taming to do initially, then it's a matter of minor color hue shifts and saturations. Lastly, use the Ha or combination of Ha, Sii and Oiii to create a Luminance layer where you can bring out the contrast and details.

Let me know if you have any questions! Looking forward to seeing some variations :-)

Extra credit: I've got it on my to do list to process an HOO (bi-color) version. Any takers on that?

Thank you Bradley, Graeme, here's what I did:

In Fitswork:
- make three 16 bit tifs.
In Photoshop:
- combine the three images as one RGB image
  (H to red, O to green, S to blue)
- use channel mixer to get the 'right' colours
- increase local contrast with Topaz Adjust
- clean up colour fringes with Topaz Denoise
  (noise reduction wasn't necessary, but
  Denoise has a few handy extras)
- tweak colours with hue/saturation
- reduce to 50% for smaller file size

This came out:

Click for a larger view   
You guys gave us some awesome source material to work on! There's a world of detail in the data and so little noise, it's amazing.

The result is a bit tacky, but I had fun so I thought I'd submit anyway. I never before did a three channel combination and I'm very grateful for the opportunity.

Had to give it a go too!
Here is my try on it. So i started with
Stretching the channels in Histogramtransform
then open the pixelmath and combined the SHO
SCNR after that
made one small curve in curvetransform
Layed tha HA as a syntetic luminancelayer to bring back some details from the ha-channel. And also to bring the noise down.
And just a touch of sharpening through Atrouswavelettransform
And the last thing i did was localhistogramequalisation.
So a very quick processing. took maybe 10min.

And maybe i should mention this process was made in Pixinsight

Here is my quick result.
Processing flow:

Stretch channels in APP
Load channels in PS
Combine channels (OIII = B, SII = R, Ha = G)
Auto color
Selective color reduce magenta
NIK-tools dfine2 (noise reduction, 40%)
NIK-tools in PS —> viveza (warmth, brightness)
SCNR in PI
NIK-tools detail extractor (30%)

Three great renditions there! They all have to be clicked to appreciate. Not sure why it's a pelican, looks more like a dragon to me! The data is quality Bradley.

Ruud wrote: ↑Sat Oct 03, 2020 12:04 pm Thank you Bradley, Graeme, here's what I did:

Nice mostly red image Ruud. The yellow highlights and the filament detail in the dark nebulosity are excellent. Is the Sii to blue channel low or was that your choice of mix? I haven't played with it yet!

Micke187 wrote: ↑Sat Oct 03, 2020 4:33 pm Had to give it a go too!

Glad you did! Interesting mix of more blue and very little red. As an NB newbie myself "Layed the HA as a syntetic luminance layer" is a bit too esoteric! I need to play!

avdhoeven wrote: ↑Sat Oct 03, 2020 9:30 pm Here is my quick result.

Great SHO, again highlighting blues, as opposed to Ruud's HOS. Is your increased range of colours a result of Selective color reduce magenta or the actions below that line in your workflow that I didn't understand!?

I best go give this a go myself now!

Regards

Graeme

Well Its easy to make a luminance layer. Just use lrgbcombination and unbox rgb and place the HA as luminance and there you go! It really makes a different on the noise. For my part i dont play too much with the saturation after. I only use stretching on the channels before combination for the colors. It usally is enough. And if the snr are low i use a rangemask and stretch only the nebulosity more before combine them but the data here was so good it wasnt really necessary.

'Oh and i forgot to mention i also made a starmask to reduce the stars a little in morphtransform.

Graeme1858 wrote: ↑Sun Oct 04, 2020 11:15 am
Great SHO, again highlighting blues, as opposed to Ruud's HOS. Is your increased range of colours a result of Selective color reduce magenta or the actions below that line in your workflow that I didn't understand!?

It is because I basically normalize the different channels to the same strength (didn't mention that in the workflow ) using levels in PS.

The Nik-tools is a very nice toolset which you can download for free for PS. There are really some nice tools in there.

I tried again.

This time I used three groups and a black layer underneath. The groups are named S, O and H, after the images they contain. Within each group there is
  a solid colour clipping layer with 100% opacity and blend mode colour
  a curves clipping layer with 100% opacity and blend mode normal
  a 16 bit image (starting off with 50% opacity for each image)

The clipping layers only influence the image in their own group

This gives a great amount of control: There is no limitation to red/green/blue channels (yellow instead of green worked out better). The curves layers add control over the contrast and brightness of each colour group, and extra freedom is added by changing the opacity of each of the three images I used 50% for the images in the O and S group and 31% for the image in the H group.

I copied the result into one final layer, did some final tweaks and reduced to half-size.

Click for full size

Hi all this is my attempt on the image,I used my combine rgb channels in anstronomy tools.I then used my light pollution software once I put it together.Then I used astronomy tools with vertical noise reduction and Horizontal noise reduction,and used lighten the dso,finally hlvg for the green and ended up with this image from the data.

Wow! Loving all of the renditions and the diversity of results! Thanks everyone!!!!

So as some have figured, the Ha image is the cleanest from a noise perspective, and that's intentional. It's the largest integration time as I always use Ha for the Luminance layer and process it separately. Ha has 26 900s exposures and Oiii/Sii has 10 900s exposures. Here's my Ha image I processed to use as Luminance. This would have the most detailed work using Deconvolution, nonlinear and linear noise reduction tools, LocalHistogramEqualization and/or HDRMultiscaleTransform, CurvesTrasnsformation and UnsharpMask.
I then recombine Ha, Sii and Oiii to obtain an RGB image on which I do color processing. I do very little noise reduction, if I do it would be nonlinear on the individual images prior to combining. I do STF applied to HistogramTransformation to stretch the RGB image.

I did the BiColor version, which is Ha to R, and Oiii to G and B using PixelMath. Prior to combining I did do a LinearFit with Ha as a reference and applying it to Oiii. This results in a stronger Oiii signal and when combined allows for more teal/blue to appear. I did a STF applied to HistorgramTransformation to stretch the RGB image. And afterwards did minor color saturation to bring blues out a little more. I always use ATrousWaveletTransform with 4 layers and all layers except the Residual disabled to blur the RGB image. Here is the RGB image before and after blurring:
I then used LRGBCombination to overlay the Luminance image obtain as above on the RGB, apply Chrominance Noise Reduction and applying saturation by lowering it from 0.5 to 0.34. Then I do more color work using masks and ColorSaturation/CurvesTransformation. I did some SCNR work and a DarkStructureEnhance at 20%. This is my result:
Hope to see some more renditions! If anyone has questions, do let me know. Would be glad to help :-)

andrewsscope wrote: ↑Sun Oct 04, 2020 6:01 pm Hi all this is my attempt on the image,I used my combine rgb channels in anstronomy tools.I then used my light pollution software once I put it together.Then I used astronomy tools with vertical noise reduction and Horizontal noise reduction,and used lighten the dso,finally hlvg for the green and ended up with this image from the data.

Very vibrant colours Andy. And a whole set of different colours again! Shows what can be done.

Regards

Graeme

Here's my one. I had a few goes at it, this is the first half decent version!

I opened all three files as layers in Gimp. Gimp actually opened 14 frames of something for each file. Don't know what that was all about so I just deleted them all except for the one with the three layers in it.

Converted to RGB from Greyscale and stretched the curves of each of them a few times each, clicking the visibility on and off as I went along to see what was happening. Then I moved H to the bottom layer as it is the brightest, O to the middle and the faintest S to the top. I stretched S and O a bit more to balance them and reset the black point with Levels on each layer.

Then I stretched the red on the H, the green on the O and the blue on the S. After playing with the colour balance in favour of the colour the layer was stretched to I adjusted the the visibility of the top layer to 35% and the middle layer to 50% and finished with an unsharp mask.



Then I did it all again with red on the H, green on the S and blue on the O.



Not brilliant but lots of fun!

Thanks Bradley.

Regards

Graeme

bradleybcraig wrote: ↑Mon Oct 05, 2020 12:01 am I always use ATrousWaveletTransform with 4 layers and all layers except the Residual disabled to blur the RGB image. Here is the RGB image before and after blurring:

What is the reason you blur the RGB before layering the luminance on top of it?

Regards

Graeme

Graeme1858 wrote: ↑Tue Oct 06, 2020 11:50 am

bradleybcraig wrote: ↑Mon Oct 05, 2020 12:01 am I always use ATrousWaveletTransform with 4 layers and all layers except the Residual disabled to blur the RGB image. Here is the RGB image before and after blurring:

What is the reason you blur the RGB before layering the luminance on top of it?

Regards

Graeme

I think its to reduce noise before he apply the luminance(HA) that is much more cleaner then the 03 channel is. After you get a more cleaner final image.

Cheers Micke

That makes sense. (Even if I can't see how it works!)

I was going to have another go using Oiii to blue and Sii to red and green with the Ha as a luminance, just for the craic!

I'll try blurring the RGB before adding the luminance to see what happens.

Regards

Graeme

Graeme1858 wrote: ↑Tue Oct 06, 2020 11:50 am What is the reason you blur the RGB before layering the luminance on top of it?

That's a good question. First I can say I do it because I saw someone else do it somewhere, can't recall where, in a video tutorial somewhere. Here's an example where you can see the results. It is equivalent to doing chrominance noise reduction. I made two previews from the main RGB image. One is not blurred and the other is. I also took the equivalent preview from the Luminance image and then applied it to each version of the color image, with no saturation or chrominance noise reduction in LRGBCombination. The left image is the non-blurred and the right is the blurred. You can see the difference in the chrominance noise between the two. You could probably achieve the same results using ACDNR on the chrominance layer on the left image before or after overlaying Luminance.

I did a little bit on it, stretch the data beforehand with a back ground extraction on each panel,scnr green reduction, psf ,deconvolution 40 iterations,rgb working space,localHistogramEqualization, hue shift on the magenta stars .tvgdenoise,crop.

i could of denoise the rgb channels beforehand to reduce the pitted look ..

i think the oiii chanel needs another star alighnment with local distortion correction, there seems to be coma shift on this panel ,maybe from imaging on the other side of the meridian or collimation.. i didn't worry about it .