L-(L)RGB Combines

[Juan Conejero]

[RGB]RGB sounds like a no-op. What we are denoting as [RGB] here is just the CIE L* component (from the CIE L*a*b* or CIE L*c*h* color spaces).

The CIE L*a*b* components (or Lab for short) of a pixel are (nonlinear) functions of its individual RGB components. The literature is plenty of descriptions about Lab<->RGB conversions, as well as descriptions about the Lab space. The same for the Lch space. Bruce Lindbloom is a great source of information on color theory: http://www.brucelindbloom.com/

Lab and Lch provide a perfect (in the mathematical sense) separation between luminance and chrominance; this is precisely their key advantage. The CIE L* component is what we usually (and often informally) call luminance. The a* and b* components define the chrominance (a* is the green-red ratio and b* is the blue-yellow ratio), as well as c* and h* (c* stands for colorfulness, or color saturation, and h* is the hue). In PixInsight, many tools perform luminance/chrominance separations on the fly, using the Lab, Lch and CIE XYZ spaces (the latter for linear images).

So a [RGB]RGB "combination" looks like a transformation pair:

RGB->Lab, Lab->RGB

which is clearly a no-op. I guess that the actual reason why someone thinks this is of any value is:

RGB->Lab, <some black box transf. applied to L>, Lab->RGB

where the black box transformation is "something" that a particular software does behind the scenes, and without asking for the user's permission. I can't see any other reason why two successive transformations between color spaces (one being the inverse of the other) can do anything but wasting time.

Now regarding a [LRGB]-RGB combination, it is, in my opinion, a contradiction in terms. If the purpose of shooting an unfiltered luminance is to improve the signal (because the RGB data have been acquired with less exposure time and/or less spatial resolution), then why re-insert some noise from the RGB data in the luminance of the final result?

Then we could talk also about LRGB and whether it is a good idea or not, but that's a different (and longer) story.

[mmirot]

Juan,

First, thanks for the PI 1.5 release. I have had only a few minutes to try 1.5 last night. The new RT wavelet module is very cool.

Yes,

It is confusing. It would seem that some of what is on this web site is incorrect or misleading.  Yes, I don't see how there is anything to a [RGB]- RGB image other than a Lab color. Perhaps, I should contact the author for clarification. He appears to have professional astronomy and imaging processing credentials.

I would like to hear your opinion of LRGB.  I going toward dropping the L and only acquiring RGB.
 It is a lot simplier to acquire and process.
My feeling is that LRGB is a poor  short cut developed by imagers when the hobby was new to amatuer ( last month) .
 It is hard to match the quality of the L to RGB data. One or the other can easily suffer.

Max

[Jack Harvey]

From my earlier post.

We have gotten away from shooting Lum frames and just use unbinned R,G,B

[Juan Conejero]

My feeling is that LRGB is a poor  short cut developed by imagers when the hobby was new to amatuer ( last month) .
 It is hard to match the quality of the L to RGB data. One or the other can easily suffer.

I wouldn't say it better. In my opinion, LRGB only makes sense with binned RGB data. Binning RGB can save a lot of acquisition time, but at the cost of decreasing spatial resolution in the chrominance. In general, I disagree with the idea that spatial resolution isn't important for the chrominance. That depends on the subject. Small-scale luminance structures require proper chrominance support (e.g. little HII regions on a galaxy).

On the other hand, more luminance inevitably leads to less chrominance. In other words, by increasing the luminance we are decreasing color saturation. Of course, we can artificially raise the CIE a and CIE b components, or lower CIE L, to match or adapt them, but then the price is (did I mention there's always a price?): noise.

Jack's quote summarizes what I think about this: shoot RGB, plus narrow band data, when appropriate.

Of course, this is just my point of view. Others can (will, surely) disagree, and I am always ready to hear different opinions and to stand corrected.

[mmirot]

This is no fun. We all agree .

Actually, I think even binning the color to increase signal to noise is over rated. 
It only helps if your sub exposure are not shot noise limited.
Most of us are now shooting long enough subframes to make this no longer a factor. 
You have to have excellent dark skies or be taking fairly short subs to have binning 2x count for much.



Max
 

[mmirot]

I wrote Mischa about his [RGB]- RGB images. You are right he hit a black box goody in PS. It appear to undergone noise reduction. See corresponence:

 I am trying to find out how you made [RGB]- RGB images. I thought most
> image processing software create a synthetic luminance when combining
> RGB.
>
> How is this different?

to be honest... i have no idea. obviously it depends on how an RGB image is 'calculated' inside photoshop ( i tired only photoshop ). if one just adds R+G+B and uses this as a luminance channel, then one gets a significantly better result. i was quite surprised myself to see this happen, but i do not know the answer as i have no insight into the mathematics of photoshop algorithms.

best

mischa