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

HIP Color Calibration

dslr astrophotography
  • Please log in to reply
3 replies to this topic

#1 BQ Octantis

BQ Octantis

    Surveyor 1

  • *****
  • topic starter
  • Posts: 1,874
  • Joined: 29 Apr 2017
  • Loc: Red Centre, Oz

Posted 25 January 2020 - 10:39 PM

G'day all,

 

Getting linear image data out of RawTherapee has opened up a few new experimentation dimensions for me, and color balance has been on the agenda in recent days. I've been familiar with G2V color balance for several years, but I was simply using G2V stars to set the white point. Unfortunately, outside of a table of Hipparcos catalog G2V stars, there's not a good database out there of G2V stars. So I found myself clicking around my images looking for stars with B-V=0.69. But it dawned on me that with linear data, I can use any star (with unclipped highlights) whose class I know—as long as I have a reference color.

 

I had come across Landon Noll's stellar classification table some time ago, which has RGB colors for every class of star out there. And chances are, there's a HIP star in almost every one of my images—the wider the field, the more to choose from. In my shot of Orion a few nights ago through my Nifty Fifty, I took exposures of 30 seconds, 10 seconds, and 1 second—so I could even see how the color offsets looked at lower levels. I picked HIP 26477, which was unclipped in all exposures. It's a Class B3V star, with a reference RGB color of (165, 185, 255). The brightest RGB combination in the set was (251, 248, 250). Clearly, the red was severely overdriven, while the green was a little less. With 14 RGB points across the three exposures, the linear correction curves were revealed in Excel—notably the trend curves intersected pretty close to 0:

 

Screen Shot 2020-01-26 at 12.30.56 PM.gif

 

Linear correction curves are easy to apply in Photoshop with Curves. I just adjusted the peak output of the Red and Green channels; the peak value for the channel was the slope of the color's trend curve × 255.

 

Screen Shot 2020-01-26 at 11.00.47 AM.jpg

 

The black point was a little trickier; I simply used Lodriguss's method for light pollution remove documented in his article in Sky and Telescope. In the article, he claims it doesn't matter whether you apply this before or after stretching; I applied it afterward. Here's the final output (with HIP 26477 marked, just under the Running Man Nebula):

 

orion_hip.jpg

 

So now I have a way to calibrate my targets for better color accuracy. At the very least, my star colors look a fair bit better…

 

Cheers,

 

BQ



#2 sharkmelley

sharkmelley

    Mercury-Atlas

  • *****
  • Posts: 2,815
  • Joined: 19 Feb 2013

Posted 26 January 2020 - 04:09 AM

The ability to produce linear data from RawTherapee certainly opens up plenty of possibilities for accurate and consistent processing of colour images.  I think using a B3V blue star as a white reference is a bit odd but everyone is free to do what they like.

I have a couple of points to add that may be helpful:

  • In Landon Noll's star classification table it is not clear which colour space gamma is used for the RGB values of the various star types. If the gamma is non-linear (which it almost certainly is) then those ratios are not appropriate for the white reference of linear image data.
  • The linear image file produced by RawTherapee already has a white reference which is probably "Daylight" around 5200K.  To use a different star type (e.g. B3V) as a white reference then you need to apply the RGB ratios that transform from the RawTherapee's white reference to the B3V white reference.

Jerry's assertion in the Sky and Telescope article that light pollution can be subtracted form linear or non-linear images is unfortunately untrue if you wish to maintain consistency of colour throughout the image.  Light pollution should be subtracted from linear data.  Subtraction of light pollution from non-linear data causes increased colour saturation and colour biases.

Mark


Edited by sharkmelley, 26 January 2020 - 04:10 AM.


#3 BQ Octantis

BQ Octantis

    Surveyor 1

  • *****
  • topic starter
  • Posts: 1,874
  • Joined: 29 Apr 2017
  • Loc: Red Centre, Oz

Posted 26 January 2020 - 05:30 AM

The ability to produce linear data from RawTherapee certainly opens up plenty of possibilities for accurate and consistent processing of colour images.  I think using a B3V blue star as a white reference is a bit odd but everyone is free to do what they like.

I have a couple of points to add that may be helpful:

  • In Landon Noll's star classification table it is not clear which colour space gamma is used for the RGB values of the various star types. If the gamma is non-linear (which it almost certainly is) then those ratios are not appropriate for the white reference of linear image data.
  • The linear image file produced by RawTherapee already has a white reference which is probably "Daylight" around 5200K.  To use a different star type (e.g. B3V) as a white reference then you need to apply the RGB ratios that transform from the RawTherapee's white reference to the B3V white reference.

Jerry's assertion in the Sky and Telescope article that light pollution can be subtracted form linear or non-linear images is unfortunately untrue if you wish to maintain consistency of colour throughout the image.  Light pollution should be subtracted from linear data.  Subtraction of light pollution from non-linear data causes increased colour saturation and colour biases.

Mark

 

Oh, I agree with the white point used for debayering—although as with Photoshop, my choices in RawTherapee are Daylight and Camera (which was Daylight in the camera). And they produce different results. I've always picked "Camera" for AP, but in terrestrial photography, I've always found the in-camera Daylight to be a bit warm. But which is one is correct?

 

Regardless, the point of the calibration isn't so much to correct the debayer process; it's to correct for the optical chain from target to sensor (assuming the sensor data is linear and accurate). For lunar processing, my color correction depends on the position of the moon in the sky. Since the moon is largely gray, auto white balance works fine; but for the rest of the sky, having a handy RGB reference and linear data makes color correction a snap. (Planetary, however, remains a PITA…)

 

Finally, with linear data, should I be setting the bottom of the curves to the black point (i.e., noise pollution floor RGB value) instead of (0,0)?

 

BQ


Edited by BQ Octantis, 26 January 2020 - 05:34 AM.


#4 sharkmelley

sharkmelley

    Mercury-Atlas

  • *****
  • Posts: 2,815
  • Joined: 19 Feb 2013

Posted 26 January 2020 - 07:28 AM

Oh, I agree with the white point used for debayering—although as with Photoshop, my choices in RawTherapee are Daylight and Camera (which was Daylight in the camera). And they produce different results. I've always picked "Camera" for AP, but in terrestrial photography, I've always found the in-camera Daylight to be a bit warm. But which is one is correct?

In both Photoshop and RawTherapee, the "Daylight" setting will often give different results from the "Camera" setting even though the camera was set to Daylight white balance.  This is because of the way the underlying Photoshop/RawTherapee colour engine works. The "Camera" setting should give the right result.

 

 

Finally, with linear data, should I be setting the bottom of the curves to the black point (i.e., noise pollution floor RGB value) instead of (0,0)?

In Photoshop with linear data, the background subtraction and the adjustment to the white balance can be done in either order.  But Photoshop does not have an easy method for performing a subtraction.  I generally create a "New Fill Layer" with the mode set to "Subtract".

 

Mark


Edited by sharkmelley, 26 January 2020 - 07:30 AM.



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





Also tagged with one or more of these keywords: dslr, astrophotography



Cloudy Nights LLC
Cloudy Nights Sponsor: Astronomics