Triangulum Galaxy (Messier 33 - Ha-RGB-Composite)

13 November 2022

This is my first attempt to create a Ha-RGB composite of a galaxy. The Ha data was captured back in October during a full moon phase using the Optolong L-Extreme filter. The color data was taken at the next opportunity which sadly was a whole month later. While initially I was thrilled by the idea of combining narrowband and broadband data it turned out a lot harder than I initially thought.

I'd like to sketch the steps I did to come to the final image:

1. Stack both datasets (Ha and RGB) separately as I would normally do.
2. Crop, DBE both images.
3. Perform a photometric color calibration of the RBG data.
4. Star-align both images.
5. Remove the stars from the narrowband image.
6. Split the narrowband image into its RGB components and discard all but the R-Channel which contains the Ha-data.
7. Remove the stars from the broadband image and keep the stars and the starless images as separate images.
8. Split the starless image into its RBG components.
9. Use the red channel of the broadband image to perform a continuum subtraction of the Ha data.
10. Combine the four images (broadband RGB + Ha without continuum) into the final starless image.
11. Continue image processing as if it was a normal broadband image.

On the subject of continuum subtraction: I can recommend a tutorial by Frank Sackenheim (german). However, instead of subtracting one image from the other, I multiplied the Ha-data with the R-channel which in my case led to a better, more distinct result:

f * r * ha

Where f is an arbitrary factor (in my case 1.4) which I used to get the intensity I wanted.

For the final combination, I did not add the Ha-data to the R-channel (I tried but didn't like the results) but instead, I subtracted the Ha-data from the blue and green channels which lead to a more pleasing image:

r_final = r_channel;
g_final = g_channel - fg * ha;
b_final = b_channel - fb * ha;

fg and fb are factors that determine how strongly the Ha-data should influence the respective channels. In my case, fg was set to 0.2 and fb to 0.25.

On the final image, however, the Ha-data was a little too subtle so I took the Ha-data as a layer into Affinity Photo added a recolor adjustment and some curves, and used the blend mode screen to emphasize the Ha regions.