M81 - but with added Halpha, using techniques from Pix Insight course. Since several commented on the process, I thought I'd expand a little.

Start the process as normal. Combine your RGB images, crop to remove areas not fully covered by all filters, remove gradients with DBE if you need to, neutralise the backround, colour calibrate, and remove noise in the background using your favourite technique (TGVDenoise, ACDNR, though I find it easiest to use Atrous Wavelets).

Now you would like to add Ha to the mix. The technique should try to maintain star colours and avoid adding noise to the background.

To preserve star colours, you only want to add Ha where there is Ha signal greater higher than the continuum emission already recorded with your R filter. To do this, you need to subtract the continuum filter from the Ha and see what is left - the "Clean Ha" signal.

Step 1: Prepare the CleanHa by calculating "Ha - (R-med(R)) * F" in PixelMath, where R is your red frame. Subtracting the med(R) ensures that the resulting signal is not clipped at zero, and only subtracts signals higher than the median. The factor F should be adjusted by trial and error to take out the continuum signal. If your CCD has the same efficiency across the R band, then a good approximation for F is:

F = (Ha filter width)/(R filter width) * (Ha sub length)/(R sub length) * (Alpha Binning)^2/ (R Binning)^2

In my case, the Ha filter is 7nm, the red about 120nm, I used identical binning, but the Ha Sub is 900 seconds, and the R subs 600s, so I get

F = 7/120 * 900/600 * (2/2)^2 = 7/80 = 0.0875

Playing with the value though, I found 0.13 to be the best number to take out just the right amount of continuum emission.

Step 2: Prevent noise in the Ha background. To do this, you need to prepare a mask so that only the areas of strong Ha signal are added, and secondly, you can use it to smooth the Ha in weak areas. I created my HaMask by stretching a clone of the CleanHa, then convolving with PSF of size 20, then I tinkered with the blackpoint and mid point sliders to get high values where there was Ha, and close to zero where there was little signal. I then used the inverse of this mask to denoise the weak Ha signal areas in the CleanHa image.

Step 3: Add the Ha back to the R image. Again using Pixel Math:

R = $T + (CleanHa - med(CleanHa)/F * Boost

G = $T

B = $T

Again, the med(CleanHa) ensures only signal greater than the median is added, dividing by F scales the Ha to match the continuum emission, and Boost allows you to decide how much additional signal you want. In my case, I found 0.5 sufficient.

You then proceed as normal - stretching the image, using HDRMultiscale transform to emphasise the key structures, combining with your deconvolved, denoised and stretched Lum image, boosting saturation (with a Lum mask), applying local histogram to make the image pop, take out the green with SCNR, then final histogram adjustments to adjust black point.

I hope this longer explanation is a little easier to follow - if not, send me a PM.

Colin