Raw frames courtesy of Deep Sky West Remote Observatory in New Mexico, USA. (deepskywest.com) Data obtained with FSQ 106EDXiii / QSI683wsg / Lodestar / Paramount MyT.

27 hrs total integration (8x300s R, 8x300s G, 8x300s B, 18x1800s H, 17x1800s O, 15x1800s S). The palette is RGB with added Ha excess in all of the RGB channels, and O-III excess added to G & B channels. No field star reduction was applied.

The previous image in the CFH palette showed that there was a ton of O-III in the Crescent Nebula. So I feel justified with the amount of O-III cyan I added to the RGB data.

Processing in PixInsight.

This is a best attempt to produce a more realistic color rendition. Ionized Oxygen (O-III) and H-beta are both approximately cyan (blue-green). And of course, H-alpha adds to the deep red. But the red really shouldn't become merthiolate red, as so often happens. It ought to be more of a salmon color due to the blending with the cyan H-beta. H-beta should be roughly 15-20% of H-alpha.

As for field star reduction... this data has been superbly imaged by DSW, and the field stars, while numerous, are not quite overwhelming. So often in my own images taken from my rather poor site, the star images are not only numerous, but badly bloated. Hence they really become a strong distraction.

That isn't so much the case here. This image is sort of borderline. Yet when I apply my field star reduction, the image looks naked. So, in this case, I left the field stars alone.

---------

Revision B -- what the heck? Tone down the red clouds, and go ahead and reduce the field stars.

There's the salmon red that I was talking about. In fact, all my references tell me that the H-alpha / H-beta line strength ratio is 2.86. So the H-a excess should be added in ratios 1:0.35:0.35 to RGB. When I do that, result is not materially different from Rev B with 1:0.2:0.2 ratios. But it is leaning more strongly toward the salmon red color.

The region around 34 Cyg shows excess O-III emission, which is quite interesting. That also shows up quite clearly in the previous CFH palette image.

[ Ahh... I see that 34 Cyg, aka P Cyg, is a well known peculiar star - a blue hypergiant. That bluish stuff probably isn't our O-III line after all. ( https://en.wikipedia.org/wiki/P_Cygni )

( https://www.aavso.org/vsots_pcyg ) ]

And there also appears, in both images, to be a faint broad swath of O-III emission along the right (West?) cloudy region of the images. And if you look very closely and carefully, you'll see that the O-III extends faintly in an arc from that west side, over the northern regions too.

---------------

---------------

Rev C -- a more accurate (?) color rendition. By carefully applying the correct ratios of OIII and Ha excesses into the R, G, and B channels, atop the observed RGB data, we can now better separate H-beta from OIII. And some surprising things pop out at us...

* Star P Cyg (34 Cyg) now looks blue, surrounded by a bit of ionized Oxygen.

* There is a tiny bubble nebula feature in the OIII just West of (to the right) of P Cyg, on the right side of that H-alpha column. And there are interesting nebular column features in that nebular cloud.

* The Crescent Nebula now looks a bit more yellow, since OIII really is more green than blue. And blended with red H-alpha, that becomes yellow. And too, the H-alpha clouds look more salmon red with H-beta being more blue than green, mixing with red H-alpha.

* The large diffuse OIII arc is now plainly visible.

(This image has had field star reduction applied)

How I did this:

1. Compute the H-alpha excess by subtracting a portion of the R channel from the H channel until the stars disappear. What's left is the excess of H-alpha over the continuum red spectrum.

HaExcess = H - k * Max(0, (R - Med(R)))

2. Compute the OIII excess by subtracting off the proper combination of G and B from the O channel. That proper combination is derived from knowing the RGB combination needed to represent the OIII line color.

There are actually two OIII lines, very close to each other: the primary line at 500.7 nm, and a secondary line at 495.9 nm. These are so close to each other that I used the color for 497 nm = RGB(0, 255, 181). More green than blue.

(where did I get that? right here - https://academo.org/demos/wavelength-to-colour-relationship/ )

So now, if B = a * O, and G = b * O, then O = (a * B + b * G) / (a * a + b * b)

Subtract some multiple of that combination of B & G until the O channel stars disappear. What's left is the OIII excess.

3. H-beta is 0.35 (= 1/2.87) * H-alpha. And the color of H-beta at 486.1 nm is RGB(0, 239, 255). More blue than green, by just a bit.

4. Now form the RGB color image from some portion ka of Ha-excess, and ko of OIII-excess:

let HaX = Max(0, (HaExcess - Med(HaExcess))),

OX = Max(0, (O3Excess - Med(O3Excess))),

then

R' = R + ka * HaX

G' = G + 0.35 * ka * HaX * (239/255) + ko * OX

B' = B + 0.35 * ka * HaX + ko * OX * (181/255)

These, R' G' B' are the required color planes for the above Rev C image. Calibrate the zero point and scaling, then stretch to visibility.

==========================

Rev D -- a careful reworking of Rev C. Using the same colorization method, but this time we properly balance the amount of H-alpha and O-III excess that we add to the RGB base image.

Field star reduction has been applied in the final image.

So how do you know which R,G,B planes to use in the color excess equations above? And which H, and O?

I decided that each channel is essentially independent from each other, and not properly balanced in any way until you combine them into an RGB display image and perform the basic color calibration. Only then will each plane have the proper amount of DC offset removed and aligned to the other color planes, and then properly scaled so that the brightest stars have a proper color balance. This is the so-called black and white color balancing. PixInsight uses an intelligent process for this, and most definitely *not* the typical histogram matching that you find in other image processing apps.

So I created an RGB image from the R, G, and B planes, and then properly color calibrated in the linear domain. Then splitting out the resulting scaled R, G, B planes to become my reference color planes.

I did the same for an HOS image, also color calibrating it in the same way. From that I know that my stars look correct, and hence the image planes will have been offset and scaled as needed.

But the median offset in the HOS image will not necessarily be the median offset in the RGB image. So we have to make them the same. And for that, I use a preview, the same one used as a ROI during the color calibration of both images. I compute the Median of that ROI for both images, then offset whichever one has the lesser median value by that difference. Now both color images have the same base statistics. I do this before extracting the individual color planes for use in computing the excesses.

... and the Median used in the above color excess equations is that reference median value, not the median of the entire image. The difference should be small, but all the images really only match inside that ROI, and so that is the median I use, call it M, for all calculations.

Now that we have proper matching in the base level statistics, the amount of R channel needed to remove the stars in the H plane tells you how to scale the H-alpha excess to fit the RGB image. And similarly for the OIII excess.

By performing these preparatory operations the final combined image will show a proper balance according to what data had been collected at the telescope. So what you see above in Rev D is as close as possible to the "true" color observed in the sky.

Sadly, this diminishes interesting features like the blueness of P Cyg, and the O-III clouds are not as prominent as before, and that little bubble nebula in the SouthEast is no longer as obvious. Now that you know it is there, you can still see it if you try.

But this now becomes a reference color image for basic truth. Of course we can, and probably will, make some images with OIII exaggerated, as with the Rev C image.

And, BTW,... take a look at that little bubble nebula in the Hi-Res version of this image. It is clearly there when you view the details. And it sure looks like a shock wavefront, just like you see in movies of bombs exploding in air.