This project was made possible by my friend, Sławek (known as diver on astropolis.pl), who send me his Ha, SII and RGB data and allowed me to process and publish it. Thank you!

It's possible and rather easy to subtract continuum light from narrowband data. Sometimes this process reveals unexpected and spectacular structures (vide: M31 Ha swirl in the core).

But is it possible to subtract narrowband (Ha, SII or OIII) data from continuum (R, G, B)? This question has been nagging at me for some time and now I've got the answer.
Yes, yes it is, as it turns out.

This process seems to make the most sense in such cases as IC405, where Ha/SII signal is dominant in the red part of the spectrum and is flooding weak signal from the reflection nebula. Likewise for any object with strong OIII emission component and weak B/G reflection component.

Thus, subtracting  Ha/SII light left me with nearly pure R signal from the reflection nebula.
There's no apparent  OIII there to leak into G and B. I suspect that any "OIII" that some people manage to register is just the continuum.

Combining it with G and B I obtained... IC405 with next to no emission nebulosity. Nearly pure reflection nebula was left.
Flaming Star without the flames, if you will.

Here's how it was done:

There's more narrowband signal in R than just Ha. SII is also very prominent in IC405 and it goes right through the R filter.

Here's narrowband Ha:


And SII:


And this is just R. The problem is obvious:


So first I had to combine Ha and SII.
I did that in the simplest way possible, just averaging the two images:

Pixelmath: (Ha + SII) /2 - let's call the resulting image "HS":



Next, I used the classic continuum subtraction Pixelmath formula, but put Narrowband in place of Continuum and vice versa. I had to iteratively find the right "f" factor but it took no more than 10 minutes. The formula is this:

Pixelmath: R - f * ( HS - med ( HS ) )
where
f = 0,28

It gave me this result:


Next up was nearly standard RGB work, except that Photometric/Spectrophotometric Color Calibration wasn't going to work - there are no stars. So I neutralized the background and equalized the histogram in PI before exporting it to PS to give it some final makeup.

For comparison, Leaving Ha/SII in the red channel and doing straight RGB combo from the same data set gives the following result:



So yes, I've got rid of most of the IC405 on this image of IC405 :-)