Colour Calibration

[narrowbandpaul]

Hi all,

I don't do a lot of RGB imaging as funnily enough it always seems to be full moon when it's clear so normally it's emission line stuff.

Recently I was playing with some data from a cloudy nights member and some of the RGB that I have taken in the past.

I understand using a G2V for colour balance, however I am puzzled by this choice. I guess it is chosen as it is a fairly typical star and so if that is defined as white then there will be a good range if stars both bluer and redder than it, which will look nice.

However, astronomically, white is not defined as being G2, instead the colourless star is Vega, an A0V. Since this is the definition of white in astronomy why don't people use this as the colour reference.

I have tried using hotter stars than G2 in a bid to replicate the A0 by picking the bluest unsaturated stars by matching stars in the UCAC4 catalogue.

But even this doesn't give the colours I would normally see (ie, striking blue spiral arms etc). Even the G2 star didn't give the normal colours.

To calculate the colour weights I used the aperture tool in maxim which adds up the total flux from the star and subtracts the background. Basically aperture photometry. I would think this is a fairly accurate way to calculate colour weights.

I have tried excalibrator which seems fairly easy to use, but hasn't given any results yet.

What I did use yesterday was a variant of this. Plate solve your image, drop it in to aladin, add the SDSS or UCAC or whatever catalogue you want. This overlays the catalogue data on to your image. In UCAC there is B and V magnitudes, but not B-V (colour index). However creating a column of B-V in aladin is dead easy. You can then sort through to find the B-V nearest 0 (for A0) or around 0.6 for G2. Then go back to maxim and use the aperture method on that star to calculate the weights. In this way, any B-V can be used to calculate the weights.

By using a star in the image surely this takes in to account atmospheric extinction to give accurate and faithful colour rendition. The results were very similar to previous attempts.

Overlaying catalogue data on your image is really cool though and really simple!

So what other methods exist for accurate colour calibration? A piece of white paper?

Another thing: if this using the aperture tool and a star in the image is actually a good way of doing it, then does that mean that images showing very different colour are in a small part fictitious?

I'm just a bit unsure really, I know how to balance colour using a G2 in the image but it doesn't seem to give the correct result....

Any thoughts...?

Cheers

Paul

[narrowbandpaul]

No one able to shed more light on the issues involved in colour calibration?

[andrewluck]

You seem to have most of the bases covered but you're making one error. G2V is the standard because it matches our local star's colour which is pretty much middle of the road. Stars such as Vega aren't white, they're at the blue end of the spectrum. Trying to use these for calibration will produce incorrect results.

I usually use Pixinsight's colour calibration which produces excellent results very simply. In the past I've also used XCalibrator with some success.

Regards

Andrew

[narrowbandpaul]

I know it's middle of the road,. It's a very average star. It is not a white star. A white star is defined in astronomy to be A0V. I know this is quite blue, but that's the definition. If it's middle of the road you want then you could ditch the G2V as the sun has peak emission at 500nm, which is bluey green. If you go for halfway in the visible spectrum (400nm-700nm) then you want a star with peak emission at 550nm. This equates to a temp of about 5300K and a K0V star with a B-V of about 0.8

So there are some stars that make sense for calibration

A0V is by definition white

G2V is a common star, but does not lie in the middle of the vis spectrum

K0V has peak emission in the middle of the vis spectrum.

Of those three, I would argue that only two have any real merit. The G2V seems to be a rather arbitrary convention don't you think?

Paul

[RikM]

I just tweak it until it looks nice

[andrewluck]

The argument for K0V would make sense if the human eye had a linear response across the visual spectrum but it doesn't. It peaks in the green (something to do with millions of years of evolution in the vicinity of a G2V star  ). To the eye, our star produces white light with an equivalent temperature of 6500K. Vega has a colour temperature of 8-10,000K. In photographic terms this is the colour of a clear blue sky.

To see this in action, take some terrestrial photos with different white balances.

So, while in astronomical terms, Vega is a white star if we were in orbit we would see it as blue.

As I said, in practice I don't worry about this and simply use Pixinsight to average the stars in the field and then use that as the colour balance.

Regards

Andrew

[narrowbandpaul]

You do raise a good point about the eye. I have compared some images with different stars used for colour balance.

An A1 star gave a more red appearance to things and it didn't look quite right. A K0V or there abouts gave a much more usual appearance.

All three have some merit it seems, with G2 -K0 giving standard results.

I tried to take an average of B-V for K0V from the All Sky catalogue of some 2.5million stars to try to get a accurate B-V to aim for. The average was about 0.82 with a spread of about 0.17. Really quite noisy results. I think a b-v of 0.7-0.8 won't go far wrong.

A boost of saturation helps too.

Cheers

Paul

[ollypenrice]

If your aim is to replicate what the eye would see in a very large and impossibly fast telescope then I guess you should work from the eye's perception, as Andrew argues. Ie the light of a G2V. I haven't done it, but I wonder if taking a series of terrestrial shots (with a serious aperture mask!) might not be the way? Find a balance which looked generally lifelike? Or image some kind of colour test card? Do we have to calibrate on the night sky? I don't know.

Olly

[RikM]

I did make a daytime white balance test a while ago. I used tin foil with a pinhole as an aperture mask and shot against a photographic 'white' reflector. It should have been a grey card but I don't have one. This returned colour weightings very close to my standard G2V values for a star more or less at the zenith.

For my G2V weightings I used the 'intensity' as measured with the aperture tool in Maxim on sets of 10x 1min averaged subs, bias subtracted, for each filter. These seem to be a good starting point (taking account of atmospheric extinction at lower altitude) but I still need to make adjustments to get the colour balance to 'look right'. e.g. my galaxies always seem to come out magenta at the start even if the stars look right.

For a quick and dirty, I use the grey point slider in the levels tool in PS CS3 and click on the background sky in a few places and you usually get one that looks 'right'. I think it could be possible to properly calibrate on the background but from most places the background would actually be red.

[ollypenrice]

I've a feeling that there may be more magenta out there than we think. In calibrated RGB images of emmision nebulae I get a magenta colour rather than a red. If I had the courage of my own convictions I'd do this;

I did post these but offered more conventional versions as well. I thnk these may be more truthful though.

Olly

[RikM]

I love that deep cherry red, very Christmasy (nothing to do with the cherry and white hoops on my favourite rugby jersey of course ) 

[narrowbandpaul]

I think the use of something like a G2-K0 provides a fairly good colour balance. The A0 was interesting but it made things look a bit red. The good thing is that in ex calibrator you can set the range of b-v to use, so you can experiment a bit to find what looks right.

The more worrying question is what IS right? There must be someway of calibrating the image so that the colours are correct.doing it by eye is fine but it lacks precision. I like being precise. I like saying this is the answer.

Olly, are these images using a convention colour calibration with no tweaks to make the image look right?

There is certainly going to be blue stuff out there, the Ha will have Hb emission as well as OIII in hotter regions. You combine the blue and the red you get magenta.

As example if you do a HaOIIIOIII image of the rosette, which as loads of Oxygen then the image comes out with a magenta core. Quite possibly in HII regions the colour might actually be magenta. After all if the gas is fully ionised then the Hb is about 1/3 of Ha and that surely gives the nebula a magenta tinge.

Cheers for the discussion, all very interesting

Paul

[ollypenrice]

Good point about the H beta bringing the reds over into magenta. Logical.

Yes, the posted images were pretty much the intital colour I detected after calibration in Pixinsight.

Olly