Style and LRGB Emulation in Bicolor

[Mike in Rancho]

As I've been doing a lot of bicolor lately - both my own L-eNhance and downloaded practice sets - I've been paying more attention to the controls and noticed that the bicolor preset in Color switches the Style and LRGB Emulation to Artistic and RGB Ratio.

Well, I suppose I've always noticed, but now I'm becoming curious about why?

It may be buried here in the forums somewhere, but in reading around and following the links and various explanations by Ivo, mostly the Artistic settings are discussed as emulating "other software," and stating or at least inferring that this is not optimal compared to ST's usual Color module settings.

Is there something inherent in the bicolor worfklow (Composite from OSC/DSLR) or the locking together of B and G in Color that makes this the appropriate option?

The obvious visual effect is the muting of color, requiring more saturation, but I have manually changed things back to constancy/scientific etc. while maintaining the bicolor matrix. It seems fine, and of course it can clearly be done, but...is that somehow wrong?

[admin]

That's a great question!

The reason for this is that, is that it is more likely that the RGB Ratio methods are better at rendering color in those cases ("those case" being any cases where luminance and color were acquired/imported separately).

For some color space manipulations to yield "correct" results, luminance and coloring are assumed to be coupled, in the sense that they are assumed to cover the exact same part of the spectrum. However, if luminance covers a different (wider, narrower, or just different etc.) spectrum response than the coloring, then those color space manipulations don't truly make sense any more.

StarTools can then fall back to a more "naive" (but possibly far more "correct") ratio preservation algorithm, where luminance does not play a role in the reconstruction of the hue and/or saturation as much (or even at all in some modes).

You should indeed not see too much difference when using data from an OSC or DSLR, since luminance and color were acquired at the same time. Slightly more dramatic effects on hue and/or saturation difference may occur when using more complex composites from different sources and/or mono cameras.

Does that help?

[Mike in Rancho]

Thanks Ivo,

It does help.

So as my photons fell through the duoband all at the same time (as you say, no crazy compositing...yet), and both the L that will be synth'ed and the RGB info captured will all be from the same narrow-ish bands, I shouldn't see much go sideways between the options other than the muting. ?

Now...just wait until I start combining it with the RGB I got on the same target...

As a follow-up on bicolor, and I know this bounces around on CN frequently including just the other night where you posted a sample screenshot, but am I right that the duoband composite of (R, GB, GB) and subsequent bicolor matrix in Color creates a teeter-totter of sorts, and one is limited to balancing between the two? (though based on how the filter works I'm not sure there'd be any other option)

Now I know that the synth L creation can be tweaked, including for bicolor from OSC (I tried it), by weighting the exposure times...but there's no analogue to that for color, correct? Whatever ST has initially read in from the dataset as the chrominance levels will stay put, other than the teeter-totter balancing mentioned?

I think that's what was being referenced by the OP of the CN thread as well -- how to boost (usually the OIII blue) without having to lose your red Ha to a light pink lol.

I am curious of that too, as it often might make a nicer picture, but...my concern is if that might start turning it into "art."

[admin]

Thanks Ivo,

It does help.

So as my photons fell through the duoband all at the same time (as you say, no crazy compositing...yet), and both the L that will be synth'ed and the RGB info captured will all be from the same narrow-ish bands, I shouldn't see much go sideways between the options other than the muting. ?

Correct.

Now...just wait until I start combining it with the RGB I got on the same target...

That's where you may see more significant differences. The biggets differences I have seen have been from LRGB datasets where L did not match the spectrum that R+G+B covered.

As a follow-up on bicolor, and I know this bounces around on CN frequently including just the other night where you posted a sample screenshot, but am I right that the duoband composite of (R, GB, GB) and subsequent bicolor matrix in Color creates a teeter-totter of sorts, and one is limited to balancing between the two? (though based on how the filter works I'm not sure there'd be any other option)

Now I know that the synth L creation can be tweaked, including for bicolor from OSC (I tried it), by weighting the exposure times...but there's no analogue to that for color, correct? Whatever ST has initially read in from the dataset as the chrominance levels will stay put, other than the teeter-totter balancing mentioned?

I think that's what was being referenced by the OP of the CN thread as well -- how to boost (usually the OIII blue) without having to lose your red Ha to a light pink lol.

I am curious of that too, as it often might make a nicer picture, but...my concern is if that might start turning it into "art."

A bi-color (in StarTools) gives you exactly what it says on the tin; two colors. E.g. two bands, perfectly mapped to two distinct hues (ideally opposing/complementary, etc. in the color wheel for optimal perception).

So, say we're talking about a HOO composite;

• If you see a white area, Ha is just as dominant as O-III

• If you see a deep red area, O-III is absent

• If you see a deep cyan area, Ha is absent

• If you see a pink area, Ha is dominant but some O-III is present

• If you see a light cyan area, O-III s dominant but some Ha is present

No other (psychovisually detectable) hues are introduced.

In a HOO example. the continuum runs like this;

Selection_719.jpg (20.98 KiB) Viewed 5825 times

Increasing the saturation narrows the continuum around the 50% mid way point so the "pastel" shades becomes fewer, while more and more vivid coloring gets mapped closer to the 50% mark.

Tweaking the color balance (via the red, green and blue bias sliders) in a HOO composite bi-color, is essentially changing the location of white area vs the 50% mark. This results in linear(!) compression or stretching the range for red (Ha) and cyan (O-III). But this is getting a bit more complicated.

[Stefan B]

Oh, that's interesting! Just had the problem that combining duo NB data as luminance and UV-IR data as RGB (all captured by a DSLR) was giving strange green colors around stars. This was not true if just processing the broadband data... I will have to try changing the LRGB emulation! Any other things I'd need to change for this kind of LRGB combination?

Regards
Stefan

[Mike in Rancho]

Thanks Ivo.

That helps a lot. But you guys get tins down there? Optolong sent me mine in a plastic. No respect.

And your description now makes a lot of sense with what I've been seeing when altering the saturation. Seems I was close to on track regarding the bicolor balance, but I'll consider in light of your continuum now too.

Some duoband stuff I see I have been somewhat skeptical of, in that they start appearing an awful lot like NB SHO (even when allegedly in HOO and not utilizing the fake SHO manipulation). There must be something else going on though. Some I can understand if it's an L-eX vs L-eN, due to the Hb line, but a lot of them are the same L-eN that I have.


Stefan that is indeed where things get interesting, and I believe where Ivo tells us we have to have a well thought out plan?

I've just charged forward to experiment before, of course. I did not have a lot of luck using even extracted "Ha" as L. It washed out my Coccon and just did not look right. I ended up combining the split with the extracted R into a newR, as explained in the Guy notes and another recent thread around here. That gave me dark red (I had to back off some) and a lot of nice additional detail.

But the last duo+RGB I tried, for the Tulip, actually utilized the NBAccent for the duo in composite. That seemed to work pretty well, with reasonably normal visual spectrum stars and nebula colors, but a big boost to both target and far-flung nebula detail that the RGB (f/s DSLR with UV-IR cut filter) just did not pick up. The separate stretch and then luminance / color controls gave a lot of options for making the combo work. Pretty utilized it as a full field ov view "accent," so perhaps a little beyond the original intent of NBA? Oops.

[Stefan B]

Originally I also tried to extract the Ha from l-eNhance and use this as luminance. But surprisingly this looked far more noisy than the original l-enhance data. Makes no sense but so I went with the original l-enhance stack for luminance. But as I've said, I had no luck with LRGB since the colors looked strange around some stars.

Processing just the broadband data and boosting it with NBaccent gave this result:

IC63,74x6min l-enhance + 80x5min UV-IR (NBaccent).jpg (501.9 KiB) Viewed 5805 times

Neither perfect nor utterly ugly. Some of the stars being red here, had green halos when processed as LRGB (don't have the image at hand). Don't know why. Have to check the LRGB emulation.

I did not have this issue before. Tried once on the flaming star nebula which turned out okay:

For IC 63 I am happy with the color and also with the detail of the ghost but I'd rather like to have the l-enhance data as luminance since it nicely pushes back the busy star field. This is not easily handeled in post (in the above image the stars are already massively reduced...30 iterations of tighten + unglow + color taming etc.). Will see what I can do

Regards
Stefan

[Stefan B]

Maybe it's not the right topic here, but it's somehow related. Here is a comparison of star colors from an RGB + NBaccent attempt and one with LRGB

RGB

RGB stars.JPG (78.09 KiB) Viewed 5755 times

LRGB

LRGB stars.JPG (94.17 KiB) Viewed 5755 times

The RGB stars look "right" to me and the LRGB stars don't. The three highlighted stars have this kind of yellowish tint with a faint red halo. The color of the corresponding RGB stars looks more uniform and natural to me. The problem didn't disappear with any of the LRGB emulation options. The color bias sliders were the same for both attempts.

I always thought that LRGB takes the luminance from L channel and color from RGB. So there shouldn't be a difference in color between LRGB and RGB but that's obviously wrong...what am I missing? Can I somehow "fix" this or is there nothing to fix?

Regards
Stefan

[Mike in Rancho]

Stefan,

All this stuff gets well over my head, so for the technical understanding we'll need Ivo to come back from resting after the new beta!

I also don't know how comparable those workflows are, NBAccent is quite a bit different since it comes in later and has its own color and luminance contribution controls.

But...when you make your LRGB, are you compositing it using the duoband as L in a pure L, RGB, or L + Synth L from the RGB also?

Also, when I test things out in similar fashion (I tried just my RGB file, duo as L, RGB, and duo as L + Synth L, RGB, but not NBA yet), I can certainly see some differences when changing LRGB emulation options. Are you also trying out the Style options? If it's not already there, putting that (back?) to Scientific rather than the Artistic choices should deepen those colors. At least it does on the data I tested.

[Stefan B]

Hi Mike,

I also don't know how comparable those workflows are, NBAccent is quite a bit different since it comes in later and has its own color and luminance contribution controls.

I had the impression that NBaccent didn't have any impact on star colors and it looked like the pure RGB stuff.

But...when you make your LRGB, are you compositing it using the duoband as L in a pure L, RGB, or L + Synth L from the RGB also?

I use pure L, RGB since I want the contrast and SNR from the duo NB filter and don't want to wash out the contrast with the RGB luminance. If I was using a real luminance filter on a mono cam with seperate RGB filters it would be a different story.

If it's not already there, putting that (back?) to Scientific rather than the Artistic choices should deepen those colors. At least it does on the data I tested.

Both screenshots came from scientific style. Changing the LRGB emulations didn't reduce the differences but rather increased them.

Regards
Stefan