HaRGB compose instructions question

[Mike in Rancho]

So I recently did an HaRGB on the Cocoon nebula, and am actually pretty happy with it right now. 5 hours of RGB, and 3 hours of L'eNhance. I followed the extract/compose instructions which are in multiple places including the manual I believe. But it got me to thinking (always dangerous ).

So first I extracted the R from the filter data and saved that as Ha. Then I loaded the RGB set into R, and the Ha into G, and saved that as my new R. Then I loaded everything up and here's where I have questions. While new R went into R, per the instructions original RGB went into G, B, and L.

I presume that doing this creates a synthetic L out of the entire original RGB? But we are calling that real L for now. But also, with the default composite setting as L+synthetic L (was this wrong?), will it not create yet another synthetic L, this time from (newR, RGB, RGB), and add it to the L, which is synthetic also as noted? I think I need to get out a yellow pad and trace this, but I was wondering if there's a weighting issue here.

Nonetheless, the extra detail added to the nebula was great. It was very, very, very deep red, even after dialing it back, but that was way better than the pink I got when I tried Ha as L.

[Carles]

HI Mike!

did you try NBAccent as well? does the R+Halpha gives you better results than RGB+NBAccent (l-enhance) ??
I am testing this too with Tulip Nebula, have gotten 4h40min RGB at Bortle7, so not great, but around 2h from L-extreme

Tried quiclky with NBAccent and it gives a big punch of nebulosity to the image but will try also splitting it and addit to R

Another question, adding it to yours, I was checking the stack from L-extreme in Siril, and watching the channels, It looks like
i get more information at Green channel than Blue.. but.. wasn't OIII was supposed to be located at Blue channel ?

So if I split l-extreme data... should I extract Oiii from Green channel ??
here's a few screenshots to ilustrate.
Red channel

Siril Red.jpg (274.11 KiB) Viewed 3979 times

Green channel

Siril Green.jpg (244.15 KiB) Viewed 3979 times

Blue channel

Siril Blue.jpg (209.54 KiB) Viewed 3979 times

Carles.

[Mike in Rancho]

Hi Carles,

I have not yet tried NB Accent. My initial understanding was that is was to accent RGB with like "poke-through" Ha spots in galactic arms and the like. I guess I need to learn what it really does and the way it can be used.

But I'm using HaRGB in my case and ignoring any OIII in the B and G because of my extraction. Funny too, I actually have a small debate going on right now with a PI guy about proper bicolor from duoband. Not sure I understand it all, but I am not following the logic of "it doesn't matter" and just leave things as RGB because the filter on its own turns that into HOO.

For your green vs blue - well, for one thing there are 2x G pixels in the bayer matrix than B (though I think Ivo has some kind of square root formula for that?). It will also depend on your filter and the spectral response of your sensor's bayer matrix. My L'eNhance is wider, but your narrower L'eXtreme is more centered around actual OIII. In a lot of discrete color filters, such as used for R, G, and B with mono cams, the OIII falls just within the G filter, but outside the B.

Here, it's likely that your OIII is falling onto G and B in some ratio relative to one another. If you are intending to do bicolor, to me the answer is to load the file into each channel using compose and pick bicolor from OSC/DSLR, then in color select the bicolor preset (of which there are then many bicolor mappings to choose from).

For just that, no separate extractions should be necessary. Unless you are intending to do something more creative like combine with other full spectrum data?

[Carles]

Hi Mike,

indeed I'm trying to do R+alpha,G,B+oiii rendition, in the same way i did here
viewtopic.php?p=10331#p10331

and, NBAccent can do much more than simple "poke" of h-alpha or Oiii.
For galaxies, that's probably it, adding touches of H-alpha, but for nebula, when using l-enhance or extreme,
the change can be dramatic.
Like what i'm working with now, which is the Tulip; 4h40min L-pro(RGB) and 2h40min L-extreme.
in this case is only Compose LRGB(Lpro) NBaccent (Lextreme) , Wipe-> Color> Shrink>SuperStruc> NBAccent. (just to make it fast)
This first screenshot is in the module but showing the Before

Screenshot-1.jpg (300.71 KiB) Viewed 3946 times

And next one is the After

Screenshot-2.jpg (288.1 KiB) Viewed 3946 times

it does bring up more nebulosity, faint one actually, all depending on how you use the first step in the module,
like whch sort of ROI and ignore fine detail values you give.
But that was already pushing or stretching a bit the RGB, if I stretch less, in order to bring less noise, and add NBAccent..
Then we got Before

Screenshot-3.jpg (292.43 KiB) Viewed 3946 times

and After.... in another post cuz max is 3 files

[Carles]

...and after

Screenshot-4.jpg (278.92 KiB) Viewed 3946 times

Did zoom at the Tulip to make more visible the difference..
not sure to what point is a good idea to use this more than once (do iterations of the same NBAccent.)
if I move the Band Balance to the right, so more Oiii I get this

Screenshot-5.jpg (253.43 KiB) Viewed 3946 times

and keeping this and do anotehr iteration, result is this one ;

Screenshot-6.jpg (290.18 KiB) Viewed 3946 times

it kept the Oiii (quite suttle in this object, and with this "short" exposure from l-extreme)
and added h-alpha. (that is , using last two presets, adecuate for dual-tri band filters.

My intention is to ad the h-alpha to Red, and Oiii to Blue, and then even reinforce it with NBAccent.

will keep trying

And about the just load it as RGB and it will get automatically to HOO...is kinda true, for what I've tested.
if you change identity in Colour module, it does a good job with SHO or HOO, bicolor.

As as I know, for the technical sheet of my sensor, it has a bayer of R, Gr, B,Gb. so the green always get some
red and some blue anyway. not sure to what point is actually 2 greens and 1 red 1 blue.. when each green has
actually some of the other too.


Carles.

[Mike in Rancho]

Wow, Carles, great explanation and examples.

I will give it a shot. Still about 10 days for me to come up with my final on the Cocoon, so I'll compare the things I can do in NB with the versions I came up with doing HaRGB and Ha as L.

All the bayer response graphs I've seen, though they may vary by camera, have some bleed over between the colors. They aren't discrete like a set of RGB filters. So, duoband on OSC isn't going to be pure Ha in the red, or pure OIII in the green and blue. That said I still try to do it as a legit and proper bicolor, especially since Ivo built it right into ST for us! I have done a few tests on my West Veil, and if I load it as RGB and oversaturate the colors heavily just to see what's there, all sorts of extra colors come out when there should just be shades of two.

Anyway, I am still curious about the Ha + R instructions in the manual (and elsewhere) and if it's still weighted okay with the various synthetics + synthetics going on. Might need to flag down Ivo or perhaps ask in the notes/Guy section.

[Carles]

Wow, Carles, great explanation and examples.

I will give it a shot. Still about 10 days for me to come up with my final on the Cocoon, so I'll compare the things I can do in NB with the versions I came up with doing HaRGB and Ha as L.

All the bayer response graphs I've seen, though they may vary by camera, have some bleed over between the colors. They aren't discrete like a set of RGB filters. So, duoband on OSC isn't going to be pure Ha in the red, or pure OIII in the green and blue. That said I still try to do it as a legit and proper bicolor, especially since Ivo built it right into ST for us! I have done a few tests on my West Veil, and if I load it as RGB and oversaturate the colors heavily just to see what's there, all sorts of extra colors come out when there should just be shades of two.

Anyway, I am still curious about the Ha + R instructions in the manual (and elsewhere) and if it's still weighted okay with the various synthetics + synthetics going on. Might need to flag down Ivo or perhaps ask in the notes/Guy section.

Thank you, Mike!

Can't wait to see your results /comparisions!

Well, not sure how the filters are applied on the pixels in the sensor. Theoretically, it should be a filter colour for each filter, acting in fact as a discrete filter for a whole sensor, but on a pixel level.

As far I've seen in the datasheet would be
Gr B
R Gb

qhy8L sensor Bayer.jpg (141.58 KiB) Viewed 3887 times

So, duoband on OSC isn't going to be pure Ha in the red, or pure OIII in the green and blue. That said I still try to do it as a legit and proper bicolor, especially since Ivo built it right into ST for us!

well, yes and No; it will be Ha and Oiii, it is just that half of the pixels will get less information as I understand. but still, the Red and Gr will receive H-alpha signal, and B and Gb will get h-beta and Oiii (in your case with l-enhance) and just Oiii in my case.

Funny thing though, using the script of "Extract Ha and Oiii" in siril, it does extract well the Oiii (guessing some part of the Gb and the B.
as you can see here .)

oiii bkg extr..jpg (234.81 KiB) Viewed 3887 times

So perhaps in this case at least, seeing how the signal is distributed in the channels /colours, maybe is better extract and stack the channels straight away, instead of made the separation in Compose module (@admin kinda need your input here! ) at least, seeein the representation of them in my previous post, where the Blue channel had almost no signal, and almost all the Oiii captured by my l-extreme went to Green channel for some reason. that said though, is true that Oiii is not purely blue, but rather greenish blue.. so maybe that's why it got more shifted to the green channel, or rather, this one was more sensitive for it ?
not an expert on the matter, so will be glad to be corrected

Regards

Carles.

[Mike in Rancho]

Hi Carles,

Well, I'm not so sure the Gr Gb in your graphs means what you think it does. To me it looks like it's just the readout direction, not anything to do with how discrete the color bayer filters are. But I'd be curious if it is. Most bayer graphs I am aware of are not discrete.

One would have to look up the QE graph or spectral response of whatever that model camera is. I can't find a factory one for my D5300 (the sensor of which is also used in a certain QHY model), but Mark Shelly made a spectral response graph for it - stock UV-IR cut anyway. I overlayed it here with the Optolong L'eNhance graph.

D5300 and L-Enhance overlay 800px.jpg (203.99 KiB) Viewed 3872 times

In any event, and regardless of bayer bleed-over, I am still in favor of compositing duoband in ST using the bicolor from OSC/DSLR setting, and then using a bicolor matrix in Color.

I'm not sure where extraction would come in for OIII - as that's almost always an Ha thing as far as I've seen, so that it can be used as L or to add to R.

I need to read up more on NB Accent too - I presume if you choose the duoband option it will treat B and G similarly to the bicolor from OSC/DSLR compose?