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

How to balance RGB in Pixinsight

  • Please log in to reply
19 replies to this topic

#1 TopherTheME

TopherTheME

    Viking 1

  • *****
  • topic starter
  • Posts: 610
  • Joined: 11 Feb 2011
  • Loc: SE Michigan

Posted 30 September 2020 - 10:20 PM

I'm in the process of learning Pixinsight, been doing a lot of reading, watching a lot of youtube, but I'm having an issue I just can't find a solution to. I'm trying to combine separate RGB monochrome images of the Iris Nebula together using the LRGBCombine tool in PI and am getting horrible results. For some reason the R channel shows up with a lot lower intensity than the G and B channels even though the image data appears to be rather equal for all of them. Here is the image and histogram with a simple equal mid-tone stretch. 

 

Image.JPG

 

Histogram.JPG

 

All of the channel weights are equal for the LRGBCombination tool and I tried doing BackgroundNeutralization and that just makes everything rainbow colored. I looked through a number of references on this process and they all assume the three channels will come out relatively equal in intensity. What could be causing this and how do I fix it? 

 

HW is a QSI 683-WSG, Astrodon LRGB filters, number of exposures (56) and duration of exposures (300 sec) are all the same for all 3 images. 


Edited by TopherTheME, 30 September 2020 - 10:23 PM.


#2 georgian82

georgian82

    Viking 1

  • *****
  • Posts: 851
  • Joined: 16 Jan 2016
  • Loc: Healdsburg, California

Posted 30 September 2020 - 10:39 PM

I could be wrong but I think you have to use Linear fit to match the intensity of all channels.

Pick the brightest channel (usually L) by looking at the histogram and then open linear fit and choose that channel in the drop down menu. Then simply drag the triangle in the lower left to the other channels. This will make all the channels be of equal brightness.

Try combining after this step and see what happens.

One last thing, I am not sure if linear fit has to be done in linear or non linear form for LRGB work. You may have to try both and see what happens.

Hope this helps

Edited by georgian82, 30 September 2020 - 10:40 PM.


#3 Ken Sturrock

Ken Sturrock

    Cardinal Ximenez (No one expects the Spanish Inquisition)

  • *****
  • Administrators
  • Posts: 9,337
  • Joined: 26 Sep 2009
  • Loc: Denver, CO

Posted 30 September 2020 - 11:09 PM

There are a number of ways. What I usually do is to run ABE or DBE on the linear RGB image without normalizing the output. That will get rid of the excessive color skew like you show. Once you do that, follow up with a background neutralization and a color balance. I generally use the Preview Aggregator script to make the background and white references from previews scattered across the image. Remember to make a new background reference after neutralization. Best of luck.


  • pfile, BenKolt and bobzeq25 like this

#4 pfile

pfile

    Fly Me to the Moon

  • -----
  • Posts: 5,593
  • Joined: 14 Jun 2009

Posted 30 September 2020 - 11:10 PM

you can just use BackgroundNeutralization after RGB combination. make sure the threshold parameter is set high enough to capture the brightest channel.

 

this kind of problem is very, very common. all astro images need some kind of color correction after RGB combination. you can use ColorCalibration after BN to get the colors right.

 

alternately if you are comfortable with plate-solving, you can use PhotometricColorCalibration which can perform the neutralization and color calibration in one step.

 

rob


  • Ken Sturrock and bobzeq25 like this

#5 pfile

pfile

    Fly Me to the Moon

  • -----
  • Posts: 5,593
  • Joined: 14 Jun 2009

Posted 30 September 2020 - 11:11 PM

ninja'd by ken. DBE first is a good idea if there are gradients, otherwise the color calibration will not be correct. when using DBE be sure to set the sample tolerance high enough to capture the brightest channel though.

 

rob



#6 bobzeq25

bobzeq25

    ISS

  • *****
  • Posts: 21,388
  • Joined: 27 Oct 2014

Posted 30 September 2020 - 11:13 PM

If you do an initial application of ABE to the combined RGB before LRGBCombination, that green background will pretty much disappear.

 

Then I do PhotometricColorCalibration, sometimes preceded by LinearFit.  Then stretch L and RGB.  _Then_ blend the L to the RGB with LRGBCombination.

 

Other workflows are of course possible, that is what works for me.



#7 bobzeq25

bobzeq25

    ISS

  • *****
  • Posts: 21,388
  • Joined: 27 Oct 2014

Posted 30 September 2020 - 11:16 PM

you can just use BackgroundNeutralization after RGB combination. make sure the threshold parameter is set high enough to capture the brightest channel.

 

this kind of problem is very, very common. all astro images need some kind of color correction after RGB combination. you can use ColorCalibration after BN to get the colors right.

 

alternately if you are comfortable with plate-solving, you can use PhotometricColorCalibration which can perform the neutralization and color calibration in one step.

 

rob

OK.  Just curious why you mentioned (the very useful) platesolving?   PCC does it semi-automatically, no need to do it separately.


Edited by bobzeq25, 30 September 2020 - 11:17 PM.


#8 Linwood

Linwood

    Viking 1

  • -----
  • Posts: 817
  • Joined: 09 Jul 2020
  • Loc: Ft Myers, Florida, USA

Posted 30 September 2020 - 11:23 PM

If you do an initial application of ABE to the combined RGB before LRGBCombination, that green background will pretty much disappear.

 

Then I do PhotometricColorCalibration, sometimes preceded by LinearFit.  Then stretch L and RGB.  _Then_ blend the L to the RGB with LRGBCombination.

 

Other workflows are of course possible, that is what works for me.

The Photometric Color Calibration also neutralizes the background (it doesn't remove gradients so doing that first seems right). 

Is the linear fit before needed or will PCC do that implicitly also? 



#9 noodle

noodle

    Vostok 1

  • -----
  • Posts: 112
  • Joined: 20 Feb 2006
  • Loc: Hong Kong

Posted 01 October 2020 - 02:34 AM

you  can use background neutralization to  fix this
check my demo video 

https://www.youtube....R14O0T&index=18

 

set a higher upper limit  :)



#10 Stelios

Stelios

    Voyager 1

  • *****
  • Moderators
  • Posts: 10,234
  • Joined: 04 Oct 2003
  • Loc: West Hills, CA

Posted 01 October 2020 - 02:35 AM

PCC rules, and it requires *nothing*--no BN, no LF. 

 

In most cases the relevant data can be taken from the image, and if not, rough coordinates from the Internet will suffice for the plate solve that PCC does. 


  • Fegato likes this

#11 Linwood

Linwood

    Viking 1

  • -----
  • Posts: 817
  • Joined: 09 Jul 2020
  • Loc: Ft Myers, Florida, USA

Posted 01 October 2020 - 04:50 AM

PCC rules, and it requires *nothing*--no BN, no LF. 

 

In most cases the relevant data can be taken from the image, and if not, rough coordinates from the Internet will suffice for the plate solve that PCC does. 

You can actually search by name right there in the option.

 

But you need to know the focal length and pixel size pretty precisely, AND if you did a drizzle integration multiply the focal length by the scale factor (default 2). 



#12 kathyastro

kathyastro

    Vanguard

  • *****
  • Posts: 2,262
  • Joined: 23 Dec 2016
  • Loc: Nova Scotia

Posted 01 October 2020 - 05:26 AM

There are a number of ways to balance the channels.

 

1. If you are going to apply the Screen Transfer Function as your Histogram Stretch, make sure you unlink the colour channels in STF first.

 

2. If I am going to combine colours while still linear, I use Linear Fit to balance the channels before doing the Colour Combination.

 

3. If I am going to combine colours after stretching them, I make sure I stretch all three histogram peaks to the same value, usually 25%.

 

Any of those methods will produce a pretty even colour balance.


  • Ken Sturrock, Alfredo Beltran and Marcelofig like this

#13 pfile

pfile

    Fly Me to the Moon

  • -----
  • Posts: 5,593
  • Joined: 14 Jun 2009

Posted 01 October 2020 - 07:35 PM

OK.  Just curious why you mentioned (the very useful) platesolving?   PCC does it semi-automatically, no need to do it separately.

well i said this since many times PCC fails to solve the images i'm looking at. this might be because i'm looking at a lot of different people's images for which i have to guess the image scale... when that happens i usually go to ImageSolver and try to tweak the scale/pixelsize/focal length until i can get it to solve. so if you've never platesolved an image before, or don't even know what it is, PCC might be a little daunting.

 

another issue with PCC is that there are parts of the sky that don't have good coverage in the APASS catalog. when PCC pulls photometric data from an area of the sky like this, the error it gives is pretty opaque. to an inexperienced user it looks like platesolving of their image failed, but instead the photometric data could not be "solved".

 

rob



#14 TopherTheME

TopherTheME

    Viking 1

  • *****
  • topic starter
  • Posts: 610
  • Joined: 11 Feb 2011
  • Loc: SE Michigan

Posted 01 October 2020 - 09:27 PM

First of all, thanks all the help. I did a few of the suggestions and they all got me to about the same point. For now I think I'm going to do LinearFit -> DBE -> PCC, that process seemed to work really well. I didn't see much of an improvement using Background Neutralization. 

 

Second, HOLY CRAP PCC is amazing! What a great feature!

 

Iris Fixed.JPG

 

 


  • Stelios, Ken Sturrock, Alfredo Beltran and 3 others like this

#15 Linwood

Linwood

    Viking 1

  • -----
  • Posts: 817
  • Joined: 09 Jul 2020
  • Loc: Ft Myers, Florida, USA

Posted 01 October 2020 - 09:30 PM

Second, HOLY CRAP PCC is amazing! What a great feature!

 

Yeah, I started recently and watched some tutorial that pretty much said "it almost never works", but every time I have had it failed when I double checked I had given it bad parameters.  I'm sure it can fail but it's really good.

 

What I struggle with is not screwing up the colors when I stretch; but I can always just run again.



#16 noodle

noodle

    Vostok 1

  • -----
  • Posts: 112
  • Joined: 20 Feb 2006
  • Loc: Hong Kong

Posted 01 October 2020 - 11:01 PM

First of all, thanks all the help. I did a few of the suggestions and they all got me to about the same point. For now I think I'm going to do LinearFit -> DBE -> PCC, that process seemed to work really well. I didn't see much of an improvement using Background Neutralization. 

 

Second, HOLY CRAP PCC is amazing! What a great feature!

 

attachicon.gifIris Fixed.JPG

you have to set the upper limit to match your background

BN1

 



#17 georgian82

georgian82

    Viking 1

  • *****
  • Posts: 851
  • Joined: 16 Jan 2016
  • Loc: Healdsburg, California

Posted 02 October 2020 - 10:11 PM

First of all, thanks all the help. I did a few of the suggestions and they all got me to about the same point. For now I think I'm going to do LinearFit -> DBE -> PCC, that process seemed to work really well. I didn't see much of an improvement using Background Neutralization.

Second, HOLY CRAP PCC is amazing! What a great feature!

Iris Fixed.JPG

Try using SNR (for green) after DBE and see what you get. Applying SNR will get rid of that green hue.

Edited by georgian82, 02 October 2020 - 10:17 PM.


#18 Stelios

Stelios

    Voyager 1

  • *****
  • Moderators
  • Posts: 10,234
  • Joined: 04 Oct 2003
  • Loc: West Hills, CA

Posted 03 October 2020 - 02:49 AM

Try using SNR (for green) after DBE and see what you get. Applying SNR will get rid of that green hue.

SCNR (Subtractive Chromatic Noise Reduction) rather than SNR (Signal to Noise Ratio usually). I know it was a typo, but might confuse someone. 



#19 georgian82

georgian82

    Viking 1

  • *****
  • Posts: 851
  • Joined: 16 Jan 2016
  • Loc: Healdsburg, California

Posted 03 October 2020 - 09:12 AM

SCNR (Subtractive Chromatic Noise Reduction) rather than SNR (Signal to Noise Ratio usually). I know it was a typo, but might confuse someone.


Yes, I meant to say SCNR. Thanks for correcting me.

#20 Tayson82

Tayson82

    Vostok 1

  • -----
  • Posts: 187
  • Joined: 17 May 2018
  • Loc: Poland / Wolomin

Posted 07 October 2020 - 09:36 AM

very fast way is to uncheck chain icone in STF and click radioactive button.




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