Pedro A. Sampaio avatar

M8 and M20


Hello everyone. I wanted some help and constructive criticism with my recent M8 and M20 project. Both in my processing/post-processing workflow and on ways to improve the final image.


Equipment:
  • Canon Lens 100-400mm F4.5-5.6L, using 250mm and F5.
  • Canon 7D Mk2 stock
  • Orion Atlas EQG
  • Svbony Mini 30mm Guide Scope
  • ZWO ASI120MC-S


Acquisition:
  • 461 x 60s with ISO 1600 - Total 7h41min integration
  • 60 darks
  • 50 flats
  • 50 dark flats
  • No bias
  • Captured with NINA
  • Guiding with PHD2 - Guiding for the most part was OK. RMS around 1.1 to 1.3" (image resolution is 3.37"/pixel)


Processing:
  • Stacked in DSS, with drizzle x2.
  • Processed in Pixinsight:

Pre-stretch:
1. Dynamic crop to cut the edges.
2. Automatic background extraction with Function Degree set to 1. I used ABE instead of DBE because DBE was giving me some weird background patterns. ABE was more consistent with the perceived gradient.
3. Background neutralization
4. Photometric color calibration
5. Multiscale linear transform for noide reduction > made a mask using luminance extraction and some histogram transformation to protect the brighter parts of the image. Used very mild parameters (layer 1: Threshold 1.000, amount 0.5, iterations 2; layer 2: Threshold 1.000, amount 0.5, iterations 1; layer 3: Threshold 0.500, amount 0.5, iterations 1; layer 4: Threshold 0.500, amount 0.5, iterations 1).

Stretch and post-stretch:
1. Stretchet using histogram, in a step by step basis. Also, increased a little bit the midtones of the red channel (which was lagging way behind). I tried to use Arcsinh Stretch, but it would make the stars extremely ugly, probably because it exacerbated the chromatic aberration from my lens.
2. Used Starnet to generate a star mask and a starless image.
3. With the starless image, I ran several instances of noise generator and convolution, topped with a slight decrease in saturation, to make the voids created by starnet more similar to the background noise of the image (starnet produces a starless image with lots of artifacts in place where the stars were). 
4. With the star mask, I produced a Halos mask following the steps by Adam Block.
5. Ran one instance of substitution using pixel math and the starless image, following Adam Block's Star de-emphasis article. Next, I ran two instances of MorpTransf (erosion) using a similar halos mask. I thought that combining both procedures worked best, since my starless image contained too much artifacts to use it again and again.
6. Using a proper mask (range mask - star mask to exclude stars), I used Multiscale linear transform to mildly increase detail in the nebulas (Used bias +0.050 in layers 2 and 3). 
7. Ran a scrit of Dark Structure Enhance at 0.2 intensity (standard is 0.4).
8. Using a range mask, I desaturated the background a little bit.
9. Using Curves, I slightly increased the contrast using an S-curve
10. Again with curves, I finished increasing the saturation.


Thank you for your help!
Well Written Helpful Engaging
Christoph Lichtblau avatar
Dear Pedro,

looking at your equipment, this is a stunning pic you are showing is here. I do not know what your expectations were, but my opinion is that you are very close to what is possible with an DSLR and normal photolenses. Very well done.

Cheers Christoph
Pedro A. Sampaio avatar
Thank you for the kind words, Christoph!

I'm well aware I'm very limited by my equipment right now… The star shapes are the best I could manage with my post-processing. Still, they're a bunch of rhomboid sores smile, with terrible purple and reddish fringes because of color aberration. As I said above, I wanted to use ArcSinh stretch to get a better non-linear stretch that preserved the amazing dynamic range of the center lagoon nebula, but ArcSinh would exacerbate the color aberration to a leve that was unacceptable.

Given all that, actually really liked the final picture, to be honest! Maybe I stretched it a little too much, as the center of the lagoon nebula is a bit too bright? Sometimes I think I may have tuned up the saturation just a bit too much as well.
Helpful Insightful Respectful Engaging
andrea tasselli avatar
If I may, the star field is really a little too red,  which shouldn't especially in proximity of the SSC. I find the centre of both M8 and M20 a tad too bright which tend to mute colours. But otherwise very well done.
Pedro A. Sampaio avatar
andrea tasselli:
If I may, the star field is really a little too red,  which shouldn't especially in proximity of the SSC. I find the centre of both M8 and M20 a tad too bright which tend to mute colours. But otherwise very well done.


Thank you for  your reply. The redness may be an artifact I introduced when stretching the image. As it was, I thought it needed more warm colors and stretched the red channel a little bit more, as it was lagging behind the other channels by quite a bit. I may have overdone it. That's a challenge to shoot with a stock DSLR, with its low sensitivity to red... As for the centers, I agree it's a bit birght, but I couldn't stretch the image without blowing up the core. Any idea what I could've done to counter that?
Well Written Insightful Respectful Concise
andrea tasselli avatar
I'd think a combination of MaskedStretch and CurvesTransformation should do the trick. But each image is different so hard to say without having it… Assuming that ArcsinhStretch is not working for you.
Pedro A. Sampaio avatar
Thank you! I'll try that combination.

ArcsinhStretch was good for the nebula… But terrible for the stars.

The reason seems to be that since stars become less saturated and white, all the chromatic imperfections show up. In a way, blowing up the stars to white "hides" the terrible aberrations.

Which gives me an idea: merge the stars from a regular histogram stretch with another image stretched using Arcsinh or masked.
Well Written Helpful Insightful
andrea tasselli avatar
Not sure that is going to work too well, because the arcsinh stretch results in very much diminished histogram, in terms of absolute values, compared to the default stretching. What you can do is use standard histogram stretching (HistogramStretching) but not as much as you would normally do to bring the mid-point in order to reveal all the nebula. Which means the mid-level is way to the right of where it would normally be. Then use ArcsinhStretch to bring out the colours as much as you can afford without blowing out the background noise.
Helpful Concise
UpperYarraObservatory avatar
Can’t offer a whole lot of advice, on how to improve this image but I can point out the good things though,
stars are not crunchy ( no sharp edges) 
noise removal hasn’t been overdone, so it doesn’t look plasticky.
colors aren’t over saturated, so subtle detail isn’t lost.
it’s a busy star field and shrinking the stars often leaves stringy bits between them, can’t see that here. 
if you pixel peep, you can find flaws in most images, so don’t get too stressed about it.

Ha images will have smaller stars, if you also have color frames some creative processing can give you the best of both worlds. I find starnet to be useful in this, a lot of trial and error will produce pleasing results eventually, but every image is unique and will require slightly different strategies.
Helpful Supportive
Pedro A. Sampaio avatar
Thanks for the input!

I've been meddling with starnet and starless images, and I think I worked something out with this image.

Here's the revision: https://www.astrobin.com/t4pj7l/B/

I processed the linear image as usual. Stretched a little using arcsinh stretch, used starnet to remove stars, and stretched the starless image with arcsinh stretch again. 

All in all, a good result. Starnet doesn't like my zoom lens, it leaves a lot of artifacts.

But no matter! My new WO GT71-II arrived, and with it, starnet works flawlesly!
Helpful Engaging