LRGB vs RGB, trying to create an apples to apples comparison - Forums

Topic starter

Dec 29, 2023 · Edited: Dec 29, 2023

Hi everyone,

probably this has been done a million times over already, however I wanted to give this a try. In a recent image thread where I shared my newest image (https://www.cloudyni...ns-of-ngc-6914/) a fellow forum member posted an interesting thought that is coming up from time to time especially with the advent of CMOS camera/sensor usage. He or she stated the recommendation only to shoot RGB instead of LRGB for the same total time because the object NGC 6914 was mostly an emission target with only parts being dust and reflection nebula.

I got curious about this and since I had ample data I tried my best to create and apples to apples comparison:

I created two data sets from the one data set to illustrate the effects of LRGB vs RGB on this particular target.

From the total data set I had I created:
- one set RGB with 3 hours per channel from the original source, total of 9 hours
- one set RGB with 1.5 hours per channel + 4.5h of luminance, total of 9 hours

I then processed both the exact same way for an apples to apples comparison. I extracted a synthetic luminance from the RGB with 1:1:1 ratio for all the luminance process like deconvolution and such.

The result can be seen here:

I created a gif and two revisions for full screen viewing of both results.

Based on what I see I drew my own conclusions and made a decision what I will do in the future, but I do not want to bias you. Parts of our hobby are also subjective in what someone prefers. See for yourself. Hopefully you find this helpful and enlightening. I created this separate thread so that it could be found easier by others in the future looking for material to look at.

Cheers from Germany,
Sven

Helpful Insightful Respectful Engaging

Oscar

Dec 30, 2023 · Edited: Dec 30, 2023

I like the RGB version better

George Hatfield

Dec 30, 2023 · Edited: Feb 4, 2024

I've found that almost any RGB image I process looks better, at least to me, as an LRGB. I think the main difference is in added contrast and detail. Typically, I create a synthetic luminance (integrate R, G, and B) or a super luminance, in which I also include any luminance data as well as narrow-band data if available. I create starless RGB and luminance images, combine them together (with their means fairly close), and then add the stars from the RGB image. I see no reason to shoot luminance. Just combine the R, G, and B masters and proceed.

George

Well Written Helpful Insightful Concise

Brian Puhl

IOTD Staff

Dec 30, 2023

Unfortunately, I don't think your animation does any good. The colors and details are very posterized. I think it goes without saying however, the same image can be achieved with only RGB, but it will require more integration, vice using luminance for all the details, and RGB for simple color.

Oscar

Dec 30, 2023

George Hatfield:
I see no reason to shoot luminance. Just combine the R, G, and B masters and proceed.

I second this

Sven Hoffmann

Topic starter

Dec 30, 2023

Brian Puhl:
Unfortunately, I don't think your animation does any good. The colors and details are very posterized. I think it goes without saying however, the same image can be achieved with only RGB, but it will require more integration, vice using luminance for all the details, and RGB for simple color.

I agree. I uploaded the revisions B and C for full screen viewing where you can manually blink between browser tabs

Die Launische Diva

Dec 30, 2023 · Edited: Dec 30, 2023

Judging from:

the brightness difference of the hydrogen emission areas at the top left corner and,
the minimal brightness difference north of NGC 6914,

I suspect what we are seeing here is just the influence of the background modelling done to the L data.

Both images are nice anyway

Tareq Abdulla

Jan 31, 2024

I keep thinking about this topic myself since long time because i live under Bortle 8/9 and Lum data is always bad polluted, so i feel like either i skip/cancel lum filter, or using LP filter as lum but not so excessive one, or just as many posted here by using only RGB and create synth L or even just go with RGB alone, but for something like galaxies and some clusters or even some nebulae where i feel that lum can stand out some details not mostly there with RGB then i feel like i must add lum filter, i prefer examples from Bortle 8/9 rather than example from Bortle 2-6, even Bortle 7 i feel it is better and cleaner a bit than Bortle 8/9 with LED main light anyway.

Mickaelle (kaelig)

Feb 3, 2024

Interesting comparison, from my experience, it depends, sometimes RGB is better & sometimes LRGB is better. I find that the Luminance brings better contrast but sometimes, it is too illuminated. After, the two images are nice.

Mickaelle (kaelig)

Feb 3, 2024

George Hatfield:
I've found that almost any RGB image I process looks better, at least to me, as an LRGB. I think the main difference is in added contrast and detail. Typically, I create a synthetic luminance (integrate R, G, and B) or a super luminance, in which I also include any luminance data as well as narrow-band data if available. I create starless RGB and luminance images, combine them together (with means fairly close), and then add the stars from the RGB image. I see no reason to shoot luminance. Just combine the R, G, and B masters and proceed.

George

Do you use a pixel Math Formula? How do you do that?

George Hatfield

Feb 3, 2024

Kaelig… I use the Pixinsight process LRGB combination often without chrominance noise reduction. For me, the most important step in creating an LRGB is using the GeneralisedHyperbolicStretch on all three images (starless RGB, starless luminance, and stars) to get a stretch that shows good contrast while keeping bright objects under control. I often use a GAME mask to help with the latter.

Well Written Helpful Insightful Concise

Mickaelle (kaelig)

Feb 3, 2024

Thank you George for your answer. I have never yet use the GHS script, is-it the CosmicPhoton one? How do you use it? Have you some advices for me. Thank your for your future answer.

George Hatfield

Feb 4, 2024 · Edited: Feb 4, 2024

There are numerous instructional videos on the Generalized Hyperbolic Stretch on YouTube (e.g., https://www.youtube.com/watch?v=WXY5kOFaxaI). It is now a process in Pixinsight, but I still use the script. It does take some practice, but it is, by far, the best way to stretch an image that I have found... at least for me.

George

David Cheng

Feb 4, 2024

From what I've done, I prefer RGB, especially on emission nebula.

Lum is mainly helpful on elliptical galaxy with less colour, or dust type nebula.

I found R + G + B is not equal to Lum, due to the filter design of interference type, with some wavelength (light pollution) cut.

Johannes Maximilian Möslein

Feb 4, 2024

Personally, my images would be horrible without Lum Filter. I often look at the RGB data and think it's bad but then I add the L and it adds all the contrast and fine nebulosity that RGB would need so much more integration time to pick up. I find myself dedicating more and more overall time to L, ranging from LRGB being 3:1:1:1 up to 6:1:1:1 depending on the target.

Well Written Helpful Insightful Concise Engaging Supportive

AdrianC.

Feb 4, 2024

I like the RGB a little better. As an experiment I tried using sloan g' r' i' instead of RGB and it looks ok ARP 143 Sloan test

The downside is that RGB, BVR, G',R',I', require more time, a big reason why the technique is not popular among amateurs. But times change, not so long ago people where taking RGB binned x2