PDA

View Full Version here: : How much L vs RGB?


codemonkey
26-05-2018, 10:20 AM
I've never really settled on how much L vs RGB I should get on a given target. I'm positive it depends on the target to some degree, for example areas with dim nebulosity will probably do well with more RGB.

For a while there I tried to gather equal parts L and (total) RGB. So for example I might get 1hr of L, combined with 20mins each of RGB. This seemed reasonable in an intuitive sense, but I had no real data backing this up, I just had no better plan so I went with it.

With that said, I'm starting a little experiment with the NGC 5078 area that I recently captured. Should the weather be agreeable I'm going to gather some more RGB data on the target and do some combinations of the data to see what produces the best (subjective) result for a given total integration time. I have insufficient RGB at the moment to do this experiment justice.

For now, here's an interesting comparison. This is slightly over 13hrs of luminance data combined with two different RGB sets.

The first set of RGB is literally one 4min sub per channel, so a total of 12mins. The second was all the RGB I gathered to date on this target, which was 140mins. Processing was identical. One thing worth pointing out with regard to the processing is that since the data sets were both very noisy, I blurred them (same amount for each image) before combining them with the L data.

billdan
26-05-2018, 11:36 AM
They both look identical to me Lee. It proves that Luminence is the most important factor as far as detail goes.
I guess you would only capture enough RGB just to get the colour SNR better.

Interesting experiment.

Cheers
Bill

Atmos
26-05-2018, 12:35 PM
Something else you may find interesting Lee is testing pure RGB and dropping the luminance entirely ;)

codemonkey
26-05-2018, 01:44 PM
Cheers Bill. I can see a difference, but it's certainly not night and day. As I said in the original post, I think it'll be different for different targets... if you have really dim nebulosity I think you'll get muddy, low contrast colours for example.



Yeah, that's the plan but I don't have enough RGB to do produce a remotely reasonable image yet. That particular comparison (LRGB vs RGB) will be a more difficult one to do fairly as well.

codemonkey
26-05-2018, 02:24 PM
Here's (https://www.cloudynights.com/topic/575319-lrgb-versus-rgb-a-fascinating-debate-at-the-highest-level/page-2#entry7851367) a great comparison on Cloudynights that demonstrates the value of RGB. I don't think NGC 5078 is a good target for it, but that's kinda the point.

I suspect that adding more RGB data might reveal more colour in IC 879 and not much else of significance.

RickS
26-05-2018, 03:05 PM
My general rule of thumb is to spend equal integration time on R, G, B and L. That means collecting (very) roughly the same amount of signal in colour and lum and hence balancing the overall SNR of the colour vs the lum. I don't have a mathematically rigorous basis for this but it seems like a sensible balance between the speed of luminance imaging and keeping richness of colour.

I have done straight RGB imaging and it works very nicely, but only for bright objects (unless you have unlimited time to throw at a each object...)

Cheers,
Rick.

codemonkey
26-05-2018, 03:48 PM
Rick, to confirm, you would typically do 1:1:1:1 as opposed to 3:1:1:1? I don't think I've ever put that much time into RGB.

RickS
26-05-2018, 06:16 PM
That's right, Lee. Often more/shorter subs for the lum (since they get sky limited faster) but the same integration time for lum as for each individual colour filter. I create a "super" lum with a noise weighted integration of the L/R/G/B masters and use that as the luminance for the image.

Cheers,
Rick.