PixInsight OSC help: combining RGB and L-eXtreme data

[Lee_P]

Hi SGL Hive Mind,

My current imaging project is M31 with my OSC camera. I'm close to 20 hours of RGB data (no filter), and would like to take L-eXtreme data to add in too, to get some glowing red hydrogen. I'm using PixInsight and haven't done this before, so am wondering if anyone can help; either by giving me some tips, a workflow, or link to a tutorial. My guess is that you process the RGB and L-eXtreme data separately and then combine them using some PixelMath wizardry..? I've done a bit of Googling but haven't found quite what I'm after. There are bound to be good resources out there that I'm missing!

Thanks,

-Lee

[Budgie1]

Not sure it this is the correct method but this is what I've tried this evening with an image of the Horse Head & Flame:

Processed both RGB & Duel-band images to the stage where they are both stretched, having already used DynamicAlignment and DynamicCrop to make sure the are both aligned and the same size.

Use Starnet to remove the stars in both and then used ChannelExtraction on both images.

Ditch the G & B channels for the duel-band image and use PixelMath to combine the Red channel from the RGB image with the red channel from the duel-band.

I used a 60/40 split like this: rgd-red*0.6xduel-red*0.4 to create an new red channel.

Recombine the new red channel with the original G & B channels using LRGBCombination but use the red channel as luminance as well as red. So your LRGB will be Red, Red, Green, Blue.

Now proceed with the normal processing with curves etc and finally add the original RGB stars back in with PixelMath.

The first image below is the original RGB image and the second is the HaRGB image.

See how you get on or see what others suggest as I'm still learning with this and my method is more of an experiment.

[vlaiv]

I can explain complete theory behind proper mixing of such data, and give you the math behind it, but not sure if you'll be willing to give it a go as it requires quite a bit of manual work to get things done - as it is not implemented in software as simple set of actions.

In nutshell, you need to convert both of your data sets to XYZ color space (RGB requires deriving transform matrix and for L-eXtreme we will do math model as it is narrow band), do processing and mixing in XYZ color space and then convert it back to RGB color space using some clever color substitution techniques.

This is of course way to get as accurate data blend as possible and closest colors to true colors as possible. You will not get Ha blend like people get when trying to emphasize Ha regions in galaxies for example - as that is sort of artificial coloring (emphasizing color to accentuate features).

If you want to replicate that sort of look - like people usually do with Ha data - like this:

then it is fairly simple procedure:

Stack and process your RGB data normally and stack your L-eXtreme data and extract red channel only - that will be your Ha data.

Stretch your Ha data so that Ha structures show nice and bright. Create starless version. Blend as new layer with "brighter" mode set (which means that it will blend Ha data if it is brighter than underlying data).

In fact - experiment with blend modes - one will give you nice blend if you boosted Ha layer properly.

[Lee_P]

Thanks @Budgie1 and @vlaiv, this is all great info and will definitely help me!

[adyj1]

As a follow-up question,  would this change if your NB data was collected with an l-enhance filter, it being triband?

Thanks

[vlaiv]

5 hours ago, adyj1 said:

As a follow-up question,  would this change if your NB data was collected with an l-enhance filter, it being triband?

Thanks

In this particular case, it would not if only Ha is extracted from L-eNhance and L-eXtreme filters.

If you want to extract both OIII and Ha data, then you would have problem with L-eNhance as it can't split Hb and OIII, so you would end up using both at the same time as single signal. Color of Hb and OIII is slightly different and this can have impact if you try to simulate true color (I say simulate because computer screens are not capable of showing pure spectral colors like Ha, OIII or Hb).