O.T. eXcalibrator - color combine ratio calibrator

[Jack Harvey]

I found this bit of freeware on another site and it is I think interesting.  I have not tried it yet but plan to tonite.  However I thought the members of this community might like to explore it.

Well, version 1.0 is finally ready.

eXcalibrator calibrates image color by using the SDSS or NOMAD1 database to
identify stars in your image that should be white. The program then computes
the correction factors to apply to the green and blue channels.

Testing has shown very good results when using SDSS data. The results with
NOMAD1 data are less consistent, however the program also makes manual
calibration much easier.

Here's the link to the download page
http://bf-astro.com/eXcalibrator/excalibrator.htm

Here's the online documentation (pdf) file (1.85 MB)
http://bf-astro.com/eXcalibrator/eXcalibrator.pdf

I would like to thank...
Mischa Schirmer & Bernhard Hubl
For their work and advice on using the (b-v) and (u-g) techniques for color
calibration. This led to the inspiration to write eXcalibrator.

Chris Abissi & Bernhard Hubl
For beta testing and software design contributions.

Herbert Raab
The author of Astrometrica, for routines to convert coordinates to image pixel
locations.

ST-ECF
For publishing the code for their Footprintfinder program as freeware.

The Centre de Données Astronomiques de Strasbourg (CDS)
For the development of The Aladin Sky Atlas

Hope you like it,
Bob Franke
http://bf-astro.com

[Juan Conejero]

Hi Jack,

Thanks for pointing this out. This tool is indeed interesting and looks great, however I disagree with these color calibration methods. I'll try to explain why, and also why we have followed a completely different path to design and implement the current color calibration tools in PixInsight.

If I understand it correctly, this tool is based on the ideas of Bernhard Hubl and Mischa Schirmer. This method is essentially the well-known G2 calibration, but instead of taking only G2 stars as white references, it allows using much more stars whose color indices (B-V and V-R differences for example) are accurately known from photometric catalogs. In both the traditional G2 calibration and this B-V calibration method, the goal is the same: make G2 stars white in the image, after correction for atmospheric and galactic extinction, as necessary. In other words, these methods are based on the idea that to achieve a valid color calibration on a deep-sky image, a G2 star must be represented as white. The advantage of the B-V method is that a lot of stars can be used as white references, since usually a pure G2 star (unaffected by galactic extinction) is hard to find.

Why G2 stars? G2 is the spectral type of our sun. G2 calibration is based on the idea that since the human vision system has evolved to perceive solar light as white, then a G2 star can be used as an ideal white reference. In my humble opinion, this is a wrong procedure when applied to calibrate deep-sky astronomical images.

G2 calibration works extremely well for daylight images, where the sun (which is a G2 star) is the only illuminant. But for the same reason, a G2 calibration is not applicable to deep-sky images: in a deep-sky image, no object is reflecting, in general, light coming from a G2 star. In my opinion, it is just that simple. Each single pixel in a deep-sky image accumulates a mixture of light coming from black-body radiators, line emission sources and reflection sources. G2 stars are just an infinitesimal part of the deep-sky ocean; pretending that our sun can be a valid white reference is perhaps a too anthropocentric vision

Our approach to the color calibration problem in PixInsight is different. PixInsight's color calibration tools are based on purely documentary criteria (as most of PixInsight): All objects in a deep-sky image should be represented in an unbiased way, as permitted by the instrumental limitations. A color calibration algorithm must not give more importance to some particular spectral types or emission wavelengths.

We currently support two color calibration methods for deep-sky images:

- The sum of all the light coming from a nearby galaxy (without a significant redshift) can be taken as a plausible white reference. This is because it includes all possible spectral types and all existing types of astronomical objects. Note that we are not saying that a galaxy is white —THIS is what is really different in our methods with respect to G2 and similar methods. We only say —and we don't want to say anything else— that by taking the integrated light from a nearby galaxy as white, we don't introduce a significant bias toward any particular type of deep-sky object, spectral type, or emission wavelength. This method has been devised by our team member Vicent Peris, who is currently working on a standardization of it. This work includes a catalog of good white reference galaxies for both hemispheres. According to Vicent, good candidates are galaxies with the following properties:

- Closer than 50 mpc
- Hubble classifications Sa, Sb, Sc, Scd, SBa, SBb, SBc or SBcd
- Inclination less than 60 degrees
- Integrated intrinsic intergalactic and galactic reddening < 0.5 mag in Johnson B

Here is an example with Vicent's NGC7331 image taken with Calar Alto's 3.5 meter telescope:

The RGB composite image before calibration:
http://forum-images.pixinsight.com/legacy/1.5-preview/ColorCalibration-3.jpg

After BackgroundNeutralization:
http://forum-images.pixinsight.com/legacy/1.5-preview/ColorCalibration-4.jpg

After ColorCalibration:
http://forum-images.pixinsight.com/legacy/1.5-preview/ColorCalibration-5.jpg


- As an alternative calibration method —but significantly more complex to implement and potentially less accurate—, we can take the integrated light from all stars in the image as a white reference. Again, we only try to avoid giving more importance to a particular spectral type by doing this. This method has more problems; for example we must take into account the effect of galactic extinction, which can be difficult. When interstellar extinction is not a problem, this method works extremely well for wide-field deep-sky images.


You are right: this is indeed an interesting topic. Besides its evident practical interest, it is also interesting because it easily unveils different (opposite?) concepts of astrophotography.

[Nocturnal]

This is indeed an interesting and timely subject. Well, it's always timely but in this case it is for me as I attempted to use the core of M31 as my white reference in my image. I may have done things wrong but at least that was my intent.

I'm going to revisit my M31 image again and see if I can make the colors 'pop' more.

[Juan Conejero]

Hi Sander,

Try with the whole galaxy instead of just the core. Or, at least, a section of the image comprising sufficient data from intermediate and outer regions. You want to include all stellar populations in your white reference. Let us know if this makes a difference.

[Jack Harvey]

Juan  Thanks for the enlightenment.  I think you know I am not a G2V fan and do not use G2V for my images.  I thought this was a bit different in that they used many stars and not just G2V.  The only remaining question I have is how does this differ from using the PI color calibration tool in which for white balance you use an aggregate of preview stars from an image?

I actually had not planned to use this for calibration but to get a ball park figure for the color combine ratios for different systems like the 3.5.  esp for when you use filters that are science type filters such as BVR etc.

[Juan Conejero]

The only remaining question I have is how does this differ from using the PI color calibration tool in which for white balance you use an aggregate of preview stars from an image?

The main difference is that our tool doesn't use a particular spectral type as a white reference. We collect a large number of stars and take the sum of their intensities on each color channel to compute a white reference.

Let us know if you achieve good ballparks for the 3.5!

[Simon Hicks]

Hi Juan,

This certainly is interesting and I'm trying to read it very carefully to understand it. I have found that the PI colour calibration tool and background calibration tool have really worked well on my images....for the first time....so I am a real convert.

However, I'm still not fully understanding why! 

But for the same reason, a G2 calibration is not applicable to deep-sky images: in a deep-sky image, no object is reflecting, in general, light coming from a G2 star.

But surely the G2V calibration method ONLY looks at the G2 star, it doesn't look at the deep sky image to get its calibration factors. It then adjusts the G2 star colour balance and 'drags' the rest of the image with it, i.e. applies the same calibration.

So assuming that the light from the G2 star has arrived at my sensor without being modified on the way, then this is a valid correction to the whole image?

Each single pixel in a deep-sky image accumulates a mixture of light coming from black-body radiators, line emission sources and reflection sources.

Surely a pixel that shows a G2 star is just recieving light from that G2 star? I am assuming that any nebulosity between that star and us is orders of magnitude fainter than the G2 star itself.

I repeat....I am a PI colour calibration convert.....but I just want to be able to argue with my astronomer friends and convince them why this is the way to go. 

[Juan Conejero]

Hi Simon,

The concepts of "true color" and "natural color" are illusions in deep-sky astrophotography. Such things don't exist. The main reason is that a deep-sky image represents objects far beyond the capabilities of the human vision system. For example, we have very small sensitivity to H-alpha (656.3 nm) at daylight (diurnal vision), and none at all at night. Most of the objects represented in a deep-sky image are so weak that we couldn't see their colors, regardless of spectral sensitivity. Trying to mimic the spectral response of the human vision system is clearly not the best option to make justice to the natural objects represented in a deep-sky image.

See this page for an interesting work that tries to represent spectral types as 24-bit RGB colors. The same source includes a page entitled what color is the sun?. Both are interesting references in my opinion.

If my sources are correct, the G2 type has standard color differences B-V=+0.63 and V-R=+0.53. This leads to the weights B=1, V=0.685, and R=0.42. So in terms of information representation motivated by color indices, the sun is not white (cyan?). By setting G2=white, which means B=V=R=1, we are ignoring both the spectral sensitivity of our vision system (we don't have a flat spectral response in the whole visible spectrum) and the true relative colors of stars and the rest of the objects. Using color indices as a color calibration pattern (why not), an A0 star could be represented as approximately white (B-V = 0 and V-R = +0.02), so calibrating with an A0 star would perhaps be more realistic, in physical terms, but again we'd be ignoring our vision's spectral response by doing so.

Color in deep-sky astrophotography is purely conventional. We assign colors that seem plausible as a (rough, necessarily) function of wavelength. For example, we represent Ha as red because it is at the red end of the spectrum. If we have infrared along with Ha and RGB, we may represent Ha as orange, and I as red. Or whatever else: it is just a conventional representation.

Our approach to color calibration is completely different. It is based on a different principle: maximize information contents in the representation of astronomical objects. This is what we call a documentary criterion. Of course, the information maximization principle is compatible, to some extent, with rendering plausible colors according to wavelengths (e.g., Ha=red).

[Simon Hicks]

Thank you Juan. This is much clearer now.

[mmirot]

Jack,

Tell us how the methods compare once you try a few.

Max