Style and LRGB Emulation in Bicolor

[Mike in Rancho]

Interesting. That may be the difference. I wonder how it would compare if you composited it the same way but with a synthetic also.

My guess is that the difference in wavelength coverage between what you are using for L and for coloring is causing the result.

I think Ivo has touched on this a number of times here and there (not that I can always understand it, but I'm getting better ), including here above.

Last night I played around with some NB Accent using 2.5 hours of RGB and 8 hours of L-eNhance. Altering things like the luminance and color contribution may well produce an effect that kind of mimics what you might get doing various things in composite such as yay or nay or synthetic or changing the exposure sliders.

Certain of the scheme choices would also give me either a deeper star color or a more yellow and even green, just as you are saying.

But, using the "clean" L, RGB composite like you are, that's more set in stone and no options to change it up. The luminance will not match up well with the color, either in spatial location or intensity. That's my guess anyway.

[Stefan B]

Mmmhhh...using L, RGB especially for duo NB and broadband was what Ivo recommended here

viewtopic.php?p=9532#p9532

@admin Ivo, what are your thoughts on this?

Regards
Stefan

[Mike in Rancho]

Mmmhhh...using L, RGB especially for duo NB and broadband was what Ivo recommended here

viewtopic.php?p=9532#p9532

@admin Ivo, what are your thoughts on this?

Regards
Stefan

Yes, you're right.

I did read through the module descriptions again also, just to try to ground myself. I was kind of under the impression that, at least for the constancy option, color remains as-is and not affected by anything in luminance, which is processed separately. But I imagine I have had it wrong or have been misreading and misunderstanding this all along. Maybe the L actually does have some inherent effect.

For the two examples you gave, the second was in L, RGB but the first was RGB with NBA. NBA has so many options that I would think it could change things quite a bit. I wonder what the comparison to L, RGB would look like it the other was just RGB (maybe with or without a synth L)?

I might dig up some data and try out some different combos. It would behoove me to have some kind of understanding of what I've been doing.

Edit: yeah even in the very beginning of this thread, Ivo is mentioning the coupling or decoupling of L and color. I am going to need to learn about this more. Maybe even figure out what things like color spaces actually even are.

[admin]

@Stefan B

Without delving into some deep, deep color theory (this is a great start if your're interested), it is important to understand that there are;

1. limitations to how color can be rendered
2. limitations to how color can be converted between color spaces (and their gamuts)
3. limitations to how color is perceived by a human (great video here)

There is a reason why there are multiple ways ("LRGB emulation") in StarTools to put luminance and chrominance together into a coherent image. The reason is that you cannot perfectly match luminance with color and the other way around. For example, on an sRGB calibrated screen, pure blue (e.g. blue 255 on an 8-bit screen) is ~11% as bright as white. Yet pure green (e.g. green 255 on an 8-bit screen) is ~57% as bright as full white.

In other words, there is a disconnect between how bright we perceive something (and can render something) and what color that something can have.

Or in other words, it much easier to show something bright that is pure green, than it is to show something that is pure blue (we run out of values much quicker).

So what do you do if you try to depict something that is extremely blue and extremely bright? Well...

• You can "cheat" by "borrowing" (using) brightness from the other channels in a way that people can't notice very well, by using a different color space that uses better (but usually not perfect) perceptual uniformity (for example CIELAB).
• You can use progressive desaturation ("fading to white") for values as they get brighter (historically the most used and the most 'naive').
• You can completely ignore brightness differences and accept that blue things will look darker (preserving color ratios).
• Or you can use combinations of all these things.

All these things will greatly impact how brighter values are rendered in your image. Stars are always the brightest things in your images, so you will notice differences there the most.

There are really no right or wrong answers here - one can argue pros and cons for any of the above.

Hope this helps!

[Stefan B]

Thanks, Ivo, for the detailed and extensive answer! I think I have an idea of the background now...

Maybe I will try to layer the duo NB and the RGB image and use the 'color of fg' blend mode. In my understanding one would get the detail from duo NB and color from RGB without any emulation algorithms. I'd like to see what that looks like.

Regards
Stefan