Jump to content

  •  

CNers have asked about a donation box for Cloudy Nights over the years, so here you go. Donation is not required by any means, so please enjoy your stay.

Photo

LRGB imaging - What % of time spent on each channel?

  • Please log in to reply
87 replies to this topic

#1 reddog1972

reddog1972

    Ranger 4

  • *****
  • topic starter
  • Posts: 372
  • Joined: 30 Nov 2016
  • Loc: Northern NJ

Posted 12 June 2019 - 10:56 AM

Looking at images posted over the course of time (and their respective acquisition details) I've seen two common themes of how people allocate acquisition time to each channel (all things remaining equal - gain/expsoure length, etc):

 

Option #1:

Lum - 50% of overall exposure time

R/G/B - 50% of overall exposure time (divided equally between the R, G & B channels)

 

Option #2:

L/R/G/B - all channels at 25% of overall exposure time

 

What I'm trying to understand is if there is a generally accepted "standard" of time allocation per LRGB channel, or does it matter on the target selected?  Does length of exposure time or camera choice have anything to do with it, or is it simply preference?  With limited clear skies in the Northeast, I'm hoping to maximize the little time I have to image with using the optimal acquisition technique.  I've tended to go the 50% Lum and 50% RGB (split evenly between the 3) route and I've liked my results.... 

 

Thoughts?!


Edited by reddog1972, 12 June 2019 - 10:58 AM.

  • elmiko and numchuck like this

#2 BenKolt

BenKolt

    Vanguard

  • *****
  • Posts: 2,122
  • Joined: 13 Mar 2013

Posted 12 June 2019 - 11:57 AM

reddog:

 

You're likely going to get a variety of answers on this as there are a myriad of choices.  My decisions are based on efficiently collecting luminance and color data with limited time, varying seeing and transparency conditions, etc.

 

When I image LRGB all together during the same night I tend to split it as you do, 50% L, 50% equally on RGB.  When I determine that my seeing may be poorer on a given night and I have several nights of imaging ahead, I might devote the whole time to RGB and shoot only L on a night of better seeing.  With sufficient time available, I may mix this up a bit to favor getting even more L or filling in all or part of RGB as needed.

 

Some plan out RGB based on the altitude of the object, using R at lower altitudes, then G, then B through the highest altitudes because of the varying effects of atmospheric extinction on those respective bands (I hope I didn't get that backwards!).  I don't do this but rather alternate RGB frames sequentially because I want to be able to get a collection of color frames through the same sky conditions with variable seeing and transparency.  However, I have been interested in experimenting with this to see if it will improve the overall blue data.

 

When shooting all together, I tend to alternate as LRLGLB, and this maintains the 50/50 mix of frames and collects the respective channel data under similar sky conditions during the night.  Presently, I also tend to bin the RGB at 2x2, which improves the SNR on those channels, maintaining the 1x1 binning on the luminance.

 

Best Regards,

Ben

 

 


  • schmeah, elmiko, Miguelo and 4 others like this

#3 Salacious B Crumb

Salacious B Crumb

    Viking 1

  • *****
  • Posts: 873
  • Joined: 13 Apr 2017
  • Loc: Greater Atlanta Area, GA

Posted 12 June 2019 - 12:44 PM

reddog when you look at the AstroBin you'll find out that a lot of guys who get most likes do very little RGB (1 - 2 hour per channel) as most of the data is in L. That limited RGB time seems to be enough to give these guys the needed color and great images. But of course as Ben points out, this is not the only way to do things.

 

 

- Mikko


  • elmiko likes this

#4 kathyastro

kathyastro

    Surveyor 1

  • *****
  • Posts: 1,633
  • Joined: 23 Dec 2016
  • Loc: Nova Scotia

Posted 12 June 2019 - 02:16 PM

This kind of choice is one reason why astrophotography is an art form as well as a technical discipline.  If you are getting good photos, there is no wrong way to do it.

 

My standard sequence is close to the 50%L 50% RGB scheme.  (It actually works out to 45% L and 55% RGB, but who's counting.) 

 

My reasoning is as follows: I want a minimum of 8 frames per filter so that I can use Windsorized Sigma Clipping when I stack them.  Five minutes is a good exposure time for L frames.  Sure 10 minutes would be better, but I also have to think about session length and bedtime!  I don't need as much time for RGB, especially since I bin them 2x2, so I do 2 minutes for them.  So, 8 frames of L at 5 minutes, for 40 minutes, plus 8 frames each of RGB at 2 minutes, for 48 minutes, gives me an 88 minute session.  That is a good session length for one night, especially if it is cold out.  If the target needs more, I can repeat it another night.

 

That works for me.  You have to work out what works for you.  If you get photos you are happy with, you are doing it right.


  • elmiko and numchuck like this

#5 ChrisWhite

ChrisWhite

    Fly Me to the Moon

  • *****
  • Posts: 5,562
  • Joined: 28 Feb 2015
  • Loc: Colchester, VT

Posted 12 June 2019 - 06:10 PM

I'm on the other end of the spectrum.  I do 1/3 in each channel and skip L.    I'd rather have the color information than luminance and find that my stars have more color and I am happier with the overall processing.  I take all of my RGB and make a synthetic Lum which I process for details and resolution.  When I combine I have a wonderfully matched luminance channel for my RGB. 


  • pfile, elmiko, AstroGabe and 4 others like this

#6 freestar8n

freestar8n

    Vendor - MetaGuide

  • *****
  • Vendors
  • Posts: 9,488
  • Joined: 12 Oct 2007

Posted 12 June 2019 - 06:42 PM

I think it's important to note that LRGB is one aspect of astro imaging that doesn't have an agreed upon model - so there is no theory to refer to that gives a basis for what is optimal.  And it heavily depends on how people blend in the L and what software to use.

 

I think all you can do is look at what other people have found to work best for them - including the software they use, the equipment they have, and the light pollution they have.

 

And no matter what - if you try different things and you find you lack color - then you should spend more time on color.

 

If you find you have plenty of color and want more detail - I guess do more lum and sharpen it.

 

But in all this - a perfectly valid fraction of time on L may be 0 - and you just use RGB.

 

You know that if you want color you can't spend 100% of the time on L.  But if you want luminance and color - you do get both with 100% of time on RGB - because RGB does carry luminance info.

 

Frank


  • Miguelo, happylimpet and numchuck like this

#7 Jon Rista

Jon Rista

    ISS

  • *****
  • Posts: 24,253
  • Joined: 10 Jan 2014
  • Loc: Colorado

Posted 12 June 2019 - 06:49 PM

I'm on the other end of the spectrum.  I do 1/3 in each channel and skip L.    I'd rather have the color information than luminance and find that my stars have more color and I am happier with the overall processing.  I take all of my RGB and make a synthetic Lum which I process for details and resolution.  When I combine I have a wonderfully matched luminance channel for my RGB. 

I've been doing the same lately, skipping L and just imaging 33% on each RGB channel. I actually do still get L, but I've been combining images out of just the RGB data. In part as an experiment, to see how LRGB vs. RGB compare, and what the real-world impact of combining L with RGB actually has (and it seems to depend on the target a lot just how much L may affect color in the end.)

 

For L to be particularly effective, you often need quite a lot more of it than RGB. I've done more of a 3:1 split between L and RGB. So something like 9 hours of L, and 1 hour each of RGB, as an example. Another option is to get 3 hours of L, an hour each of RGB (50/50 split), but create an optimized super-luminance out of all four channels by integrating the integrations of each together. This can actually have a marked impact on noise profile:

 

L Integration:

iekHuzf.jpg

 

Optimized Super-L (stack L, R, G and B integrations):

7iJ6XpZ.jpg

 

You then combine the RGB, then LRGB combine the super luminance with the RGB as normal. 


  • elmiko, AstroGabe, happylimpet and 2 others like this

#8 ChrisWhite

ChrisWhite

    Fly Me to the Moon

  • *****
  • Posts: 5,562
  • Joined: 28 Feb 2015
  • Loc: Colchester, VT

Posted 12 June 2019 - 08:09 PM

I've been doing the same lately, skipping L and just imaging 33% on each RGB channel. I actually do still get L, but I've been combining images out of just the RGB data. In part as an experiment, to see how LRGB vs. RGB compare, and what the real-world impact of combining L with RGB actually has (and it seems to depend on the target a lot just how much L may affect color in the end.)

For L to be particularly effective, you often need quite a lot more of it than RGB. I've done more of a 3:1 split between L and RGB. So something like 9 hours of L, and 1 hour each of RGB, as an example. Another option is to get 3 hours of L, an hour each of RGB (50/50 split), but create an optimized super-luminance out of all four channels by integrating the integrations of each together. This can actually have a marked impact on noise profile:

L Integration:
iekHuzf.jpg

Optimized Super-L (stack L, R, G and B integrations):
7iJ6XpZ.jpg

You then combine the RGB, then LRGB combine the super luminance with the RGB as normal.


Skipping lum might have an even greater benefit for those of you with heavy light pollution. Color filters have gaps between that can help cull at least some light pollution (every little bit helps) whereas lum gathers it all in.
  • pfile, Miguelo, ks__observer and 1 other like this

#9 freestar8n

freestar8n

    Vendor - MetaGuide

  • *****
  • Vendors
  • Posts: 9,488
  • Joined: 12 Oct 2007

Posted 12 June 2019 - 08:17 PM

One additional twist on this. If you find that after you balance rgb the color noise is mostly in one channel such as green, then instead of spending time on L you can spend more time on G. I would leave the subexposure the same but get more frames.

There is no need to have the color exposures balanced somehow - and there is no need to have the same number of frames. And for many people - there may not be need for L.

Remember that any time dedicated to L is time that could have gathered more color. It involves a sacrifice of color and isn’t just free photons to add to the image.

Frank
  • Miguelo, happylimpet, numchuck and 3 others like this

#10 Terry R

Terry R

    Viking 1

  • -----
  • Posts: 722
  • Joined: 31 Jan 2013
  • Loc: Australia

Posted 12 June 2019 - 09:34 PM

I grab as much Lum as possible.  This is where all the detail is.  With colour, about 8 subs per channel are enough for my system.  The resulting colour data can look pretty evil at times, but it's not that important as the details are in the Lum.   As far as a formula, experiment and record your findings.  


  • ks__observer and AIP like this

#11 DaveB

DaveB

    Apollo

  • *****
  • Posts: 1,447
  • Joined: 21 Nov 2007
  • Loc: Maryland

Posted 12 June 2019 - 09:41 PM

Another option is to get 3 hours of L, an hour each of RGB (50/50 split), but create an optimized super-luminance out of all four channels by integrating the integrations of each together.
 

...You then combine the RGB, then LRGB combine the super luminance with the RGB as normal. 

This is exactly what I do, except that I target 12h total exposure, with 6/2/2/2 hours of L/R/G/B. No scientific reason, but it has worked well for me.

 

One other important note, however, is that it does depend on the target. My primary targets are galaxies, so (for the most part) luminance details are more important to me than color saturation. That's why the heavy emphasis on L frames. But some nebulae have different requirements. My M45 was RGB only, and on a few Ha heavy areas, I have either used Ha only or Ha+L for the Lum channel. But my default is a 3:1 ratio of L:R:G:B. I will probably spend some time optimizing those ratios based on a more detailed analysis, but I have bigger fish to fry first.


  • numchuck likes this

#12 jhayes_tucson

jhayes_tucson

    Fly Me to the Moon

  • *****
  • Posts: 7,388
  • Joined: 26 Aug 2012
  • Loc: Bend, OR

Posted 12 June 2019 - 09:47 PM

Just remember that the amount of color data does not determine the "amount" of color that your image will display.  The amount of color data determines the level of noise in the color channels (the a & b channels in Lab color space.)  The trick is to gather enough color data to drive the color noise down to an acceptable level.  That's why 2x2 binning is often a good way to minimize the amount of RGB data needed to achieve an acceptable level of color noise.

 

I've done a lot of both LRGB and RGB imaging and both work well.  I also commonly use the approach that Jon mentioned of using a super-L channel composed of both actual Lum + Synthetic Lum constructed from the RGB data.  To use this approach, the RGB channels has to be gathered as 1x1 unbinned data.

 

John


  • leemr and robedude like this

#13 Jon Rista

Jon Rista

    ISS

  • *****
  • Posts: 24,253
  • Joined: 10 Jan 2014
  • Loc: Colorado

Posted 12 June 2019 - 09:53 PM

I grab as much Lum as possible.  This is where all the detail is.  With colour, about 8 subs per channel are enough for my system.  The resulting colour data can look pretty evil at times, but it's not that important as the details are in the Lum.   As far as a formula, experiment and record your findings.  

This really depends on the object. Detail is in the color as well. Lum spans the spectrum without any differentiation in details that are different colors. So microcontrast can actually be lost with a strong lum, and a heavily denoised/blurred RGB channel will not preserve details that only differ in terms of color and not intensity, so those details can be lost for good. This is one of the things I've noted while experimenting with RGB only imaging. I've had a terrible year so far and have not had the opportunity to compile the kind of demonstration I would like to actually show this...but it's on my list of things to do the next time I have clear skies (if I ever have clear skies again...)

 

Another thing here is that going super shallow on color, such as just 8 subs, may not allow you to pick up much of any color on fainter details. Galaxy halos, IFN, dark dust, etc. These things definitely have color, in fact when you see them in high enough resolution sometimes far-flung halos or tidal streams in galaxies are actually colorful streams of stars, often with a warmer (yellow-orange) color...but being as faint as they are, people often do not get deep enough color to actually show those colors. Dominated by colorless luminance, galaxy halos are often rendered in a rather bland gray...which may not actually be representative of the true color of those parts of the galaxy. 


  • AstroGabe, happylimpet and numchuck like this

#14 leemr

leemr

    Messenger

  • -----
  • Posts: 453
  • Joined: 21 Mar 2014

Posted 12 June 2019 - 10:20 PM

Here's an interesting example I posted on another forum last year.

 

A  bit over 13hrs of luminance combined with two different RGB data sets. One of those sets was a total of 3x4min  subs, one red, one green, one blue. The other was 2hrs 20mins RGB data, which was all I had at the time. Both images had noisy RGB so I blurred both of them (same amount) before combining them with the L.

 

comparison.jpg

 

And here is a great counter-example that demonstrates the worth of RGB.

 

My main point is that I think it depends on the target. For some, investing a lot in RGB won't yield much of a return, but for others it will.


  • schmeah, n2068dd, Miguelo and 2 others like this

#15 schmeah

schmeah

    Fly Me to the Moon

  • *****
  • Posts: 5,669
  • Joined: 26 Jul 2005
  • Loc: Morristown, NJ

Posted 12 June 2019 - 10:24 PM

I image in light polluted northern NJ as you do. I can tell you with no reservation, having tried many combinations including synthetic luminance and super luminance with RGB, that true luminance ( at least 50% and preferably twice as much as RGB) has been the key for me. My RGB images as a stand alone are absolute garbage without the Luminance. But of course depending on your equipment and skies this may not be the case for you ( though I strongly suspect it would ). You just have to try both and see what works.

 

Derek



#16 leemr

leemr

    Messenger

  • -----
  • Posts: 453
  • Joined: 21 Mar 2014

Posted 12 June 2019 - 10:28 PM

create an optimized super-luminance out of all four channels by integrating the integrations of each together. This can actually have a marked impact on noise profile:

That's a very interesting comparison, Jon. How did you perform the integration? I've tried creating a "super L" in the past but the super L always looked to me to be indistinguishable, or worse, than the L by itself.



#17 Jon Rista

Jon Rista

    ISS

  • *****
  • Posts: 24,253
  • Joined: 10 Jan 2014
  • Loc: Colorado

Posted 12 June 2019 - 10:41 PM

That's a very interesting comparison, Jon. How did you perform the integration? I've tried creating a "super L" in the past but the super L always looked to me to be indistinguishable, or worse, than the L by itself.

I use Juan Conejero's method in PixInsight, which is super easy. You integrate each channel normally, so you end up with four images (fits, xisf). Reset the ImageIntegration tool to defaults, bring all four files in, and integrate. That's it. 



#18 Jon Rista

Jon Rista

    ISS

  • *****
  • Posts: 24,253
  • Joined: 10 Jan 2014
  • Loc: Colorado

Posted 12 June 2019 - 10:44 PM

Here's an interesting example I posted on another forum last year.

 

A  bit over 13hrs of luminance combined with two different RGB data sets. One of those sets was a total of 3x4min  subs, one red, one green, one blue. The other was 2hrs 20mins RGB data, which was all I had at the time. Both images had noisy RGB so I blurred both of them (same amount) before combining them with the L.

 

attachicon.gif comparison.jpg

 

And here is a great counter-example that demonstrates the worth of RGB.

 

My main point is that I think it depends on the target. For some, investing a lot in RGB won't yield much of a return, but for others it will.

pbkoden's example is actually a really good example of how much difference in color you can see when you integrate just the RGB. It is indeed noisier, but there is actually a pretty marked difference in color... The dust lanes in the RGB only have that brownish-red "rusty dusty" color, which I think is more accurate, while the luminance version shows most of the dust lanes grayish. That would be the difference in the depth to which the longer RGB subs were actually picking up rarer photons from those low level emissions from the dust lanes..


  • leemr and FiremanDan like this

#19 leemr

leemr

    Messenger

  • -----
  • Posts: 453
  • Joined: 21 Mar 2014

Posted 12 June 2019 - 10:45 PM

I use Juan Conejero's method in PixInsight, which is super easy. You integrate each channel normally, so you end up with four images (fits, xisf). Reset the ImageIntegration tool to defaults, bring all four files in, and integrate. That's it. 

Wonder what the difference was, pretty sure that's exactly what I did...



#20 leemr

leemr

    Messenger

  • -----
  • Posts: 453
  • Joined: 21 Mar 2014

Posted 12 June 2019 - 10:48 PM

pbkoden's example is actually a really good example of how much difference in color you can see when you integrate just the RGB. It is indeed noisier, but there is actually a pretty marked difference in color... The dust lanes in the RGB only have that brownish-red "rusty dusty" color, which I think is more accurate, while the luminance version shows most of the dust lanes grayish. That would be the difference in the depth to which the longer RGB subs were actually picking up rarer photons from those low level emissions from the dust lanes..

Exactly my thoughts. After seeing his post, and performing my own little experiment, I decided I'd make more conscious decisions on how I do my LRGB, as I think it depends on the target. But then I promptly went back to doing what I always did...


  • Jon Rista likes this

#21 Jon Rista

Jon Rista

    ISS

  • *****
  • Posts: 24,253
  • Joined: 10 Jan 2014
  • Loc: Colorado

Posted 12 June 2019 - 10:48 PM

Wonder what the difference was, pretty sure that's exactly what I did...

I guess it might depend on just how big the difference between L and RGB is. I've done it in cases where I had quite a bit more L, and the benefit was almost non-existent when integrating it with the RGB channels. In fact, I've actually found a couple cases where creating a super luminance actually degraded the SNR rather than improved it, due to just how poor the SNR was in the RGB channels. I think when you have enough signal in each channel, such that none of the color channels drag the SNR of the combination down, then it works pretty well. 


  • leemr likes this

#22 leemr

leemr

    Messenger

  • -----
  • Posts: 453
  • Joined: 21 Mar 2014

Posted 12 June 2019 - 10:51 PM

I guess it might depend on just how big the difference between L and RGB is. I've done it in cases where I had quite a bit more L, and the benefit was almost non-existent when integrating it with the RGB channels. In fact, I've actually found a couple cases where creating a super luminance actually degraded the SNR rather than improved it, due to just how poor the SNR was in the RGB channels. I think when you have enough signal in each channel, such that none of the color channels drag the SNR of the combination down, then it works pretty well. 

That makes sense... I probably did have a reasonably significantly higher SNR in L than in RGB, as I do tend to invest more in L



#23 Jon Rista

Jon Rista

    ISS

  • *****
  • Posts: 24,253
  • Joined: 10 Jan 2014
  • Loc: Colorado

Posted 12 June 2019 - 10:53 PM

That makes sense... I probably did have a reasonably significantly higher SNR in L than in RGB, as I do tend to invest more in L

Yeah, that's what I've mostly done. I don't have many opportunities to do LRGB imaging in my back yard, and getting out to the dark site has been all but impossible the last couple of years. I will say, getting lots of L, even if it has an impact on color quality and depth, does indeed seem to produce a much higher SNR result from my back yard. And it really does depend on the target...some targets there just isn't much in the way of faint enough structures to really warrant deep RGB, others you could probably go so deep on RGB as to spend 100 hours or more on a single object and still not have enough. :p 



#24 BenKolt

BenKolt

    Vanguard

  • *****
  • Posts: 2,122
  • Joined: 13 Mar 2013

Posted 12 June 2019 - 11:59 PM

Last September I did a study of LRGB vs. RGB combinations using both 1x1 and 2x2 binning.  I posted some of the results here for those interesed.  My conclusion validated that I benefit from concentrating on 1x1 luminance data for resolution of detail and a "reasonable" amount of 2x2 RGB for chrominance.  I image from my suburban backyard.  My comparison composite images included SNR and noise measurements, which helped to quantify my conclusions.  As I said above, my decision to image LRGB as I do is mainly due to the economics of time.

 

Best Regards,

Ben


  • schmeah, coinboy1 and SeymoreStars like this

#25 freestar8n

freestar8n

    Vendor - MetaGuide

  • *****
  • Vendors
  • Posts: 9,488
  • Joined: 12 Oct 2007

Posted 13 June 2019 - 03:08 AM

The issue of LRGB vs. RGB is somewhat unique in astro imaging because it never comes up in professional imaging.  It is described as something that results in "higher SNR" - but that isn't a good description - and if it were, it would be used professionally as a way to get better data in less time.  But it isn't - and it doesn't.

 

The original motivation for LRGB was to take advantage of how the human eye-brain works in looking at images - along with the way high read noise ccd's behave when binning.  The eye is less sensitive to detail in color information - and ccd's tend to have less read noise when binned during acquisition.  If you put those two things together - you can save time and get a better looking image by combining unbinned lum with binned rgb.  But it was all based on how ccd's used to work - and how the eye still works.  It wasn't about "SNR" per se.  And it was just for amateur imaging to make a good looking picture in limited time.

 

If you do this all nowadays with low read noise cmos - the motivation to bin rgb is lost.  Instead - you just capture the rgb separately from lum - and process them separately so that you aggressively smooth the rgb and sharpen the lum - and combine them.  And the hope is that your eye won't notice the loss of detail in the color.

 

For some people this works well and is a big time saver.  For others it is less clear.

 

So there is no clear answer - because it all depends on how you process and what you want in an image.  The two results are not the same - and LRGB doesn't "get you there sooner."  Normally, if you expose rgb more and more - it will converge to a low noise rgb image.  But if you combine L with rgb - it doesn't converge to the same thing.  It just ends up looking nice in terms of color and noise.  But it isn't the same - and that's why "SNR" loses its meaning.  The results are different - and depend on how it was all done and what you want in the end.

 

Frank


  • Gregory and SeymoreStars like this


CNers have asked about a donation box for Cloudy Nights over the years, so here you go. Donation is not required by any means, so please enjoy your stay.


Recent Topics






Cloudy Nights LLC
Cloudy Nights Sponsor: Astronomics