Having problems with colour balance in SHO

StuartT

Well-known member
I have some integrated lum, S, H, O data and I am trying to make a colour image.

First I combined the linear SHO images using RGB combination (Sii->R, Ha->G, Oiii->B) and was then planning to stretch it and combine it with the stretched Lum image.

But the combined RGB image is excessively green even after PCC (seen on the right with a STF) . What am I doing wrong?

here are the master LSHO images
 

Attachments

  • Capture.JPG
    Capture.JPG
    731.7 KB · Views: 78
this is pretty normal since Ha is the strongest signal and gets mapped to the G channel.

if you unlink the channels in STF you might see something a little less green.

probably a good way to go about this is to use the old color calibration tools - first find the background area and use the largest value as the tolerance in BackgroundNeutralization. actually set it slightly higher.

then you can use ColorCalibration - turn off structure detection for the white balance, and turn on "output white reference mask". look at the background values in the image again (they will have changed) and set the white lower limit just above that value. for the background reference you can set the upper limit to the same value.

now run CC and undo/redo while fiddling with the white lower limit until you see only the nebula in the white reference mask. when that looks good then the colors should be a tad more balanced.

at this point sometimes i do SCNR to green at 100%. sometimes not, depending on how it looks. then follow this flow to get the SHO colors right:


you can also just use the unlinked STF copied to HT instead of all the color balancing stuff, that is certainly a strategy. but if you want to use MaskedStretch or ArcsinHStretch you need to get the colors balanced in the linear stage.

with GHS you can do a per-channel stretch so you might be able to get the colors looking less green from within that script.

rob
 
this is pretty normal since Ha is the strongest signal and gets mapped to the G channel.

if you unlink the channels in STF you might see something a little less green.

probably a good way to go about this is to use the old color calibration tools - first find the background area and use the largest value as the tolerance in BackgroundNeutralization. actually set it slightly higher.

then you can use ColorCalibration - turn off structure detection for the white balance, and turn on "output white reference mask". look at the background values in the image again (they will have changed) and set the white lower limit just above that value. for the background reference you can set the upper limit to the same value.

now run CC and undo/redo while fiddling with the white lower limit until you see only the nebula in the white reference mask. when that looks good then the colors should be a tad more balanced.

at this point sometimes i do SCNR to green at 100%. sometimes not, depending on how it looks. then follow this flow to get the SHO colors right:


you can also just use the unlinked STF copied to HT instead of all the color balancing stuff, that is certainly a strategy. but if you want to use MaskedStretch or ArcsinHStretch you need to get the colors balanced in the linear stage.

with GHS you can do a per-channel stretch so you might be able to get the colors looking less green from within that script.

rob
Ok, I did the background neutralization. But on the color calib, I am not sure I completely understand what you are telling me here. I measured the background at 0.0015 so this is what I have set in the White Ref lower limit and BG upper limit. Here is what I get. Changing the WB lower limit downwards produces a mask with a lot more white in it. While changing it upwards makes a mask which is nearly all black (i.e. just stars).

But in all cases the image is still green
 

Attachments

  • background.JPG
    background.JPG
    686.9 KB · Views: 36
  • higher.JPG
    higher.JPG
    497.8 KB · Views: 40
  • lower.JPG
    lower.JPG
    646.3 KB · Views: 54
How about trying one of the dynamic formulas?

The one from The Coldest Nights is a favourite:

R = (Oiii^~Oiii)*Sii + ~(Oiii^~Oiii)*Ha
G = ((Oiii*Ha)^~(Oiii*Ha))*Ha + ~((Oiii*Ha)^~(Oiii*Ha))*Oiii
B = Oiii
 
How about trying one of the dynamic formulas?

The one from The Coldest Nights is a favourite:

R = (Oiii^~Oiii)*Sii + ~(Oiii^~Oiii)*Ha
G = ((Oiii*Ha)^~(Oiii*Ha))*Ha + ~((Oiii*Ha)^~(Oiii*Ha))*Oiii
B = Oiii
Thanks. I do have that one set up as a saved process icon, as it happens. But it's not working for me on this image. Here it is after an unlinked STF (maybe that stretch isn't doing it justice?)
 

Attachments

  • Capture.JPG
    Capture.JPG
    403.8 KB · Views: 50
It needs to be applied to stretched images, is that how you are using it?

Edit I assume not given you're applying an STF :)
 
Having said that I wouldn't expect it to come out fully red though. Do you have the output image set as RGB?
 
btw - i should say that PCC may or may not work well on narrowband images, but i know people have gotten it to work. the old flow does work OK tho.

Ok, I did the background neutralization. But on the color calib, I am not sure I completely understand what you are telling me here. I measured the background at 0.0015 so this is what I have set in the White Ref lower limit and BG upper limit. Here is what I get. Changing the WB lower limit downwards produces a mask with a lot more white in it. While changing it upwards makes a mask which is nearly all black (i.e. just stars).

But in all cases the image is still green

you want the white ref to look like just the nebula and not capturing any background. the idea is to make the whole nebula the white reference so that the colors balance out to white.

it may still be somewhat green after this but the first step in rick's flow is to kill some of the green.


rob
 
It needs to be applied to stretched images, is that how you are using it?

Edit I assume not given you're applying an STF :)
Ah.. I didn't appreciate that fact. I was applying it to linear images. (I can never figure out which processes have to be applied to which type of data!)
Here it is applied to stretched images
 

Attachments

  • Capture.JPG
    Capture.JPG
    365.7 KB · Views: 58
btw - i should say that PCC may or may not work well on narrowband images, but i know people have gotten it to work. the old flow does work OK tho.



you want the white ref to look like just the nebula and not capturing any background. the idea is to make the whole nebula the white reference so that the colors balance out to white.

it may still be somewhat green after this but the first step in rick's flow is to kill some of the green.


rob
ok, I wasn't sure whether to adjust the number up or down (as I don't really understand what is going on). Getting most of the nebula also gets the stars tho. Is that a problem?
 
that's OK on the stars, it's not really avoidable. what you don't want to do is to click "structure detection" - that tries to find just the stars. you'd use that if you were trying to use CC in sort of the way that PCC works (except PCC is a lot more exact since it uses astrometry and photometry, CC is just trying to extract what it thinks might be stars and set the average to white.)

rob
 
that's OK on the stars, it's not really avoidable. what you don't want to do is to click "structure detection" - that tries to find just the stars. you'd use that if you were trying to use CC in sort of the way that PCC works (except PCC is a lot more exact since it uses astrometry and photometry, CC is just trying to extract what it thinks might be stars and set the average to white.)

rob
Thanks, Sorry for the endless questions! I assume I don't need to put anything in either of the "reference image" boxes (since I am applying the process to a target image)?
 
no - you don't. i think CC just lets you optionally color balance the target image by looking at another image.
 
Back
Top