Statue of Liberty Nebula - NGC 3576 and NGC 3603

This image really turned out much better than I had expected. Going into capturing data, I didn't have particularly high hopes. We've had incessantly cloudy weather here, which cleared up just in time for the full moon, and while I usually avoid imaging under moonlight, I couldn't resist starting a new project and testing out some new mods I've made to my 8" Newt. I convinced myself this would just be "test data" and that I couldn't expect much from a moonlit, partly-cloudy imaging session.

Soon after the data started rolling in, I realised that the imaging time wasn't going to be wasted. My new Antlia SHO filters were doing a decent job of cutting through the moonglow, and the seeing was exceptional. I'm usually pleased if I can capture a half-dozen subframes around the zenith with a FWHM under 1.8". However, PixInsight's SubframeSelector reported a median FWHM (Moffat4) around that figure across the channels for the entire data set. My best subframes were around 1.3".

This is only my second SHO image, so I'm still finding my feet with processing. My first pass with this data was a total disaster. I used every tool under the sun, combined various exotic stretches, and ended up with a result I was extremely unsatisfied with. The STF on each individual channel looked far better than the final image. The result was less than the sum of its parts.

So, I went back to basics and decided to simply try applying a manual MTF (HistogramTransformation) stretch to each channel, followed by a basic SHO combination. It looked much better, however it highlighted the deficiencies in my data - my OIII had been impacted by the moonlight, had a residual gradient, and the SNR wasn't the greatest. Meanwhile the SII data was quite weak. So I went back and collected more subframes, with a focus on SII and OIII.

In the end, processing was simple. After deconvolution, star removal and gradient removal, I manually matched the black-points and mid-tones to achieve a pleasing colour balance. Once the channels were combined, the image was virtually done. No SCNR was needed, because the green had been balanced via the stretching process. I applied HDRMultiscaleTransform to tame the highlights (while not going overboard - I still wanted the bright areas to be bright, just not clipped). Some noise reduction with NoiseXTerminator, a gentle UnsharpMask to the highlights, some final stretching and curves adjustments to add a bit of drama, final star recombination (using RGB data for colour), and that was it. In terms of the number of steps in the process, it was simple, but I was very pleased with the results.

The emphasis on SII, and the fact that colour balance was achieved purely by stretching the OIII and SII data, rather than diminishing the green Ha, means the result has very prominent SII regions. I'm happy how the subtle swirls of SII in the background "popped" in the final image, particularly in the starless version.

Variants
In addition to the "final" and starless versions of the image, I've prepared a few variants:

Drizzle 2x: I processed the image right to the final stages using drizzled data. This variant omits the final resampling. Even at the ambitious 0.39" pixel scale, I think it holds up OK.

More Prominent Stars: The star and starless layers were separately stretched. This variant stretches the star layer a bit more heavily. I prefer it when zoomed in, but on one of my monitors, it appears much less attractive when heavily downsampled in a web browser. So I opted to keep it as a variant and put the version with lighter stars as the "final" version.

Traditional Processing: There still seem to be plenty of people with doubts about AI processing. This version is for you. Instead of BlurXTerminator, it uses Richardson Lucy Deconvolution. Instead of NoiseXTerminator, it uses Mure, TGV and MMT. I'd like to say it's "AI Free", but that's not quite correct. I still used StarXTerminator for generating ringing supports and for isolating the stars for stretching and applying RGB colours. Life is too short for manually crafting star masks. This version ended up with slightly more of a Ha weighting during stretching, but I didn't mind the result, and left it as-is.

Crops
In addition to the main attractions, there are a few small details of interest.

PN G291.4-00.3

Tiny PN G291.4-00.3 stands out prominently in the starless image, but is difficult to spot in the star-filled image.


SII Areas

The emphasis on SII means there are a lot of interesting swirls of SII in the background right across the top half of the image (particularly visible in the starless version).

However, in the darker region near the top-left, there were some bright, scattered patches of SII that really stand out. I was originally convinced were a reflections or artifacts, yet they show up in all my subframes on both sides of the meridian. As more data came in, I became convinced they are real, and simply so prominent due to the dark region that surrounds them.


NGC 3603 Jets

The SHO processing makes the jets around NGC 3603 appear bright green. These jets are predominantly Ha, and even with the Ha channel somewhat de-emphasised in this combination, the Ha signal clearly stands out.