Milkyway Core & Rho Ophiuchi Cloud Complex


(Jonas Wagner) #1

I usually shoot nightscapes and not pure astro photography shots, but I liked this one quite a bit.
The processing was fairly tricky as the galactic center is fairly low in the sky around here.
I wanted to bright out the detail in the dust clouds & sharpen the image without completely blowing out the stars (at which makes the image very nervous and in my opinion unnatural).

I’d call the result just about acceptable. The stars don’t look like static and largely retain their colors. But a lot of the relative luminance differences got lost. If someone has some hints on how to do a better job at this I’d love to hear about it.

The basic processing workflow was:

  • Darktable to develop the RAWs to 16 bit tiff files
  • Hugin to align the tiffs
  • Siril to stack the exposures
  • Gimp & G’MIC & a lot of patience for processing

(Flössie) #2

Jonas, you are the expert on astro here (and your example here shows that again), so I’d like to ask you a (maybe stupid) more basic question: How do you manage to retain the colors of the stars at all? For if I’m taking a series of shots for stacking (at let’s say 16mm crop, f/2, 16s, ISO 3200), I capture a lot of the even fainter stars, but the brighter ones are all white and burnt out. Do you take your shots with bracketing, or what is the trick here?

Admiring
Flössie


(Morgan Hardwood) #3

I’d call it wonderful :slight_smile:

What is that pentagon-shaped star pattern in the lower-right?


#4

If the highlights are clipped, use two darker exposures for each brighter exposure. So, 8-second shots at the same settings, for one-stop highlight savings.


(Jonas Wagner) #5

I’m not quite sure what you mean. Could you elaborate a bit? :slight_smile:

@floessie there are a few things that I’ve found to help:
Always underexpose the shots. Not to the extent to clip the left of the histogram but a good bit. I’ve found little to no benefts in SNR when exposing to the right with my cameras but it does clip the highlights.

Getting enough shots & stacking them to get the noise down helps. For larger stacks doing a mean (preferably with some outlier rejection) outperforms the median with regards to retaining colors.

The next crucial step is denoising. Masking of the stars can help, as can careful selection of the denoising parameters.

The step I’m still struggling with is when raising the exposure & contrast of the image to bring out nebulas and other faint objects other than the stars. Quick example:

Input: (gas clouds hard to see but relative star brightness is readily apparent)

Curves & sharpening applied:

I don’t think the pentagon has a name. The big star is Antares. The clouds above form the Rho Ophiuchi cloud complex.


(Morgan Hardwood) #6

Try the Wavelet tool in RawTherapee. Not sure which exact control, but there was talk of using it for astrophotography. Searching here and in github will probably yield something.


(Flössie) #7

Jonas, thanks a lot for this suggestion. I always thought maxing out the number of photons was the way to go. Now your way of taking the shots relaxes the need for long exposure, high ISO, or fast lens a bit.

I was thinking a bit like CarVac: Take a series maxed out and another with reduced ISO. Then try to combine them. Hugin 2016.0.0 comes with a “zero-noise output sequence”, but I have no clue how to use it. Have you tried that? Could that help with the noise problem?

Anyway, that was eye-opening.

Thanks,
Flössie


(Jonas Wagner) #8

@floessie My observation has been that the ISO has little influence on the amount of noise once both pictures are raised to the same brightness (at least within reason) therefore I keep the aperture and exposure time but lower the sensitivity. So capturing the dynamic range is not really the issue for me.

@Morgan_Hardwood I tried that too. I always either ended up with ugly gradient reversals around the stars or losing sharpness in the dust & gas clouds. I’ll need to do some more experimentation. :slight_smile:


(Morgan Hardwood) #9

@Jonas_Wagner have you tried NAFE (Noise Adaptive Fuzzy Equalization)?
http://www.zam.fme.vutbr.cz/~druck/Nafe/Index.htm


(Jonas Wagner) #10

@Morgan_Hardwood no, but it does look interesting. I’ll give it a shot when I’m on windows. :slight_smile:


(Morgan Hardwood) #11

I used it in Linux through wine about a year or two ago, results were interesting.