Is a G2V calculation possible in Pix?

[LD]

Hello All,
I keep meaning to finally calculate my own color weights and was wondering if it can be done in PixInsight. I can certainly do it--once I shoot my star--in AIP4WIN but was curious about Pix since it has become my primary post-processing program
Thanks for any ideas,
Larry

PS: Posted this in Gallery earlier by mistake. Same message now moved.

[Carlos Milovic]

Hi Larry

I-m not versated in the procedure, but I think that you'll need to calculate de PSF of some sample stars (identified G2V sources, right?). So you'll need the intensity peak, the total flux, or the magnitude of the star, for each channel.
Since the PSFExtractor process has not been released yet, you'll need to use an approximate method, with the current tools. The most accurate I can think of, is to calculate the total flux of those sample stars. Just create a preview small enought to avoid other stars, of too much background, but contains the whole star. Now, use the Statistics process to get the (pixel values) average of this preview. After that, clone the preview (or move it, or create a new one), and place it over a pure background area as closest as you can to the star. Calculate de average again. Substract this value to the previoust average that you have. Now, you have the average flux of the star (just multiply by the number of pixels to the the true total flux, and then by 2^16-1 to get the ADUs). I think that this is all you need to apply the formulae that you always use.

Having said all of this, just in my humild opinion, I think that the G2V is a wrong procedure to follow. A G2V like color calibration is used in common daylight photography, and it works, becouse it is used on reflected light. We ussually use a white or neutral gray chart, illuminated by the current sunlight (the tone varies along the day, and if it is cloudy or not). Then, we use the average value of the chart for each channel to get the propper channel weights. I have been using this method for a while to use my modified Canon 300D for daylight pics with great success... and also it works with artificial light sources.
This method is based on the fact that the human brain ignores in certain amount the illumination's hue, and tends to nuetralize the colors, specially for well known objects. But light sources work in a very different way than reflected light objects. If you look at a black body (a star) of the same temperature of the sun, or a G2V, you are not going to see a white dot/ball. It is yellow. Perfectly white stars do not exist on nature... what you see is a apparent neutralization, becouse the reddish and bluish emitions are similar in the visual range, so you have a quite flat spectra. This is true for hotter stars than the sun. I may be wrong, but if my memory does not fails, it was something near F4 stars.
Vicent Peris has been using "flat spectral objects" (like some cuasar cores) to calibrate a dozen of narrow width filters to produce color images, with outstanding results. See the Alhambra project images. He now uses a more refined method, based on stars, that deals with atmospheric absortion too... just stay tuned, becouse he may upload a tutorial about this topic soon.

[Juan Conejero]

I think that the G2V is a wrong procedure

Amen

[LD]

Carlos,

Wow! Talk about upsetting the apple cart of accepted wisdom!  :lol: But I must say that it absolutely answers a nagging thought I've had since first encountering the G2V method: why are we going for "white balance" on a "yellow" star? You've addressed that one.

I also fully intend to follow-up on what you mention and look into Vicent's methods. The depth of what you guys at Pleiades do always amazes me, and I constantly find myself challenged and intrigued by the discussion on this forum. Not sure I always grasp it  :? but fun nonetheless.

Of course, I will probably still shoot my G2v just to see what I get with my filters. :wink:

Thanks again for that very informed and challenging reply!

Regards,
Larry

[Jack Harvey]

This is great stuff.  I am not a G2V user but some of the guys in our group absolutely swear by it, and swear at you if you don't use it<G>.  Great discussion.

[vicent_peris]

Carlos,

why are we going for "white balance" on a "yellow" star? You've addressed that one.

I'm sorry to say that the Sun is not yellow. It has more blue emission than red, and it has the maximum at the green wavelenghts.


Vicent.


PD: Kill G2Vers.  :lol:

[Carlos Milovic]

Vicent, you forget the human perception it is far less sensible for the blues than the greens/red, so it is perceived yellower


Edited: Just to add another point of view, take a look at the following link:
http://www.vendian.org/mncharity/dir3/starcolor/

[vicent_peris]

Vicent, you forget the human perception it is far less sensible for the blues than the greens/red, so it is perceived yellower

Yes, but you know what I'm saying...  :wink:

I think nobody wants really to follow the spectral sensitivity of the eye in the deep sky work.


Vicent.

[Carlos Milovic]

Of course not ;) it would be a "bye bye" for the H alpha emission.

Also, I forgot the atmospheric absortion, wich also makes the sun yellower at middle heights.
Anyway, we agree that it is not the best standard.

[Jack Harvey]

Vicent Peris has been using "flat spectral objects" (like some cuasar cores) to calibrate a dozen of narrow width filters to produce color images, with outstanding results. See the Alhambra project images. He now uses a more refined method, based on stars, that deals with atmospheric absortion too... just stay tuned, becouse he may upload a tutorial about this topic soon.

Any word on when we might see this?

[Jack Harvey]

An interesting thread on the CCDWare forum re G2V
   
   Posted 25 December 2007 06:24 AM
Hi Adam/Stan

just thought I would post my observations having followed the tutorial to redo a G2V calibration on my Astrodon RGB filters in an STL11K camera.

Very interesting!

I used 3 software programs to verify the result.

Both Maxim and AIP4WIN2 agree on the weights as 0.92:1:0.95 and interestingly both of these had the star flux as Green having the minimum, followed by Blue, and then Red.

However, CCDStack measured the flux as Green having the minimum, followed by Red then Blue. Consequently, the weights came out at 0.97:1:0.92...almost the exact reverse for Red and Blue when compared to the results yielded in the other two programs.

Most curious....would you have any observations? I can post or email the very small images if you would like.

cheers
Martin
    
Posts: 184 | Registered: 18 June 2005   Reply With QuoteEdit or Delete MessageReport This Post
Stan
Orbiting around Earth
   
   Posted 25 December 2007 01:43 PM    Hide Post
Hi Martin,

Load the final R, G, and B images then crop them to include the "G2" star and some sky background. Save as unsigned 16 bit FITS and post.

I suspect that the difference you observe is due to how background is handled, but I would like to see the actual data.

Stan
    
Posts: 738 | Registered: 14 November 2005   Reply With QuoteEdit or Delete MessageReport This Post
Ignored post by Stan posted 25 December 2007 01:43 PM    Show Post
Martin Pugh
Good Seeing
   
   Posted 26 December 2007 04:33 AM    Hide Post
Hi Stan

thanks

I have posted the 3 FITS files to:

www.martinpughastrophotography.id.au/Test_Page.htm

they are about 500Kb each.

I look forward to your assessment.

cheers
Martin
    
Posts: 184 | Registered: 18 June 2005   Reply With QuoteEdit or Delete MessageReport This Post
Ignored post by Martin Pugh posted 26 December 2007 04:33 AM    Show Post
Stan
Orbiting around Earth
   
   Posted 26 December 2007 09:09 AM    Hide Post
Hi Martin,

I get [0.92:1:0.96], which agrees with the other programs but not with your CCDStack attempt. Here is what I did:

1) "Color"; "Create". Set ratios to 1:1:1

2) select background area and apply

3) draw box around star and press the "Balance white/grey..." button.

4) double-click the area below the "Green" lable to normalize factors so that green=1.

Note that the star may be too bright (max > 40,000 ADU). If the star is near the non-linear zone then the ratios will be distorted. A good trick for color calibration is to de-focus the star.

Stan
    
Posts: 738 | Registered: 14 November 2005   Reply With QuoteEdit or Delete MessageReport This Post
Ignored post by Stan posted 26 December 2007 09:09 AM    Show Post
Martin Pugh
Good Seeing
   
   Posted 27 December 2007 04:37 AM    Hide Post
Hi Stan and thanks for that. I am glad that you got it to agree with the other two programs. Of course, I was just following the tutorial on Adam's DVD...and was not aware of this other procedure you documented.

Interestingly, these are supposed to be "tru-balance" Astrodon filters i.e ratio of 1:1:1. Clearly not.

cheers
Martin
    
Posts: 184 | Registered: 18 June 2005   Reply With QuoteEdit or Delete MessageReport This Post
Ignored post by Martin Pugh posted 27 December 2007 04:37 AM    Show Post
Stan
Orbiting around Earth
   
   Posted 27 December 2007 08:03 AM    Hide Post
The ratios are sensitive to sky conditions, especially sky-alt (lower alt extinguishes blue much faster than red). I really don't put much faith in rigid ratios. Feel free to balance each image for the most pleasing hues. There is no such thing as "true color"! <g>

[Juan Conejero]

The ratios are sensitive to sky conditions, especially sky-alt (lower alt extinguishes blue much faster than red). I really don't put much faith in rigid ratios. Feel free to balance each image for the most pleasing hues. There is no such thing as "true color"! <g>

That's very good!  :lol:  I agree completely!

Beyond that, IMO the problem with G2V is quite simple: deep-sky objects don't reflect sunlight. They emit or reflect light that originates in black-body emission. So trying to calibrate the colors in a whole deep-sky image by setting G2V=white (or yellow, or whatever else; the reference color isn't the actual problem) isn't a correct procedure IMHO.

Other concepts that simply don't make sense are "natural color" and "true color". We are representing emissions and objects that our eyes will never be able to see, so what's natural color? As Carlos and Vicent pointed out in other posts, you really don't want to follow the human eye's response curves to represent deep-sky subjects. Many colors that we accept as "true" in our DS images (e.g. Ha=red or OIII=cyan) are actually just conventional representations.

During 2008 I hope we'll be able to develop a method based on flat spectral responses that Vicent has applied to calibrate Alhambra images with great success. In this method, some quasars can be used as "flat" calibration sources, or even stars with known spectra and a special interpolation procedure that neutralizes spectral curves. These procedures are more complex than G2V calibration, but they are much more accurate and are not based on wrong (IMO) assumptions.

[twade]

To all,

I found a nice site that shows the colors of stars according to their RGB bit depth.  

http://isthe.com/chongo/tech/astro/HR-temp-mass-table-bymag.html

Browsing through the list it appears the "best" star for a correct white balance is F8IV.  It had values of 255 252 253.  I may try to find one tonight and see what happens.  

I've always considered G2V calibration incorrect.  If you use it, there's absolutely NO way to achieve a neutral background.   In fact, if you achieve a neutral background, you NO longer have a G2V calibrated image.

Wade

[twade]

To all,

Last night was clear enough for me to calculate some weights.  I have 50mm square Astrodon filters.  

Using a F8IV star (SAO 43875):  RGB ratio --> 0.84:1.00:1.79
Using a G2IV star (SAO 62949):  RGB ratio --> 0.95:1.00:1.59

It's quite interesting that the results would differ so much.  Looking at the table I mentioned yesterday, the color values for each star differ slightly.

G2IV  255 244 242
F8IV  255 252 253

Does anybody have any thoughts?  Now I'm curious as to which ratio will produce the best results.

Wade

[Juan Conejero]

Hi Wade,

Zenital distance may have a strong effect over measured colors on images. Did you take it into account to correct your measurements?

By the way, the site you linked in your previous post is really interesting and useful. Thanks for sharing it.

Now I'm curious as to which ratio will produce the best results.

Probably none. In my opinion, matching a star in a deep-sky image to the color that corresponds to its spectral type is a wrong calibration procedure. This is independent on the spectral type you use.

The problem here, as we discussed for standard G2V calibration, is that the objects represented on a deep-sky image are not reflecting light emitted by the star you're calibrating with. Along with that, most of the light in a DS image originates from black body emission (or real-world approximations to black bodies).