Green Stars! Help Needed

[Simon Hicks]

Hi all,

I've recently acquired a Canon 1000D and modded it with the Baader filter. I have taken a pile of 3 min subs near the Heart and Soul Nebula in full RGB through a Pentax 67 lens. However, whatever I do, I find that the smallest stars are always green....see zoomed in jepg of a central region below:



In the above image you can see that ALL of the smaller stars are green.

This is the result of 20 subs stacked in DSS....but I get exactly the same result with a single sub...just noisier.

I start with the Autosave file (32bit TIF), then use PixelMath with the following expression to neutralise things...

R   $T[0]/med($T[0])
G   $T[1]/med($T[1])
B   $T[2]/med($T[2])

By the way, the starting medians for the RGB are 0.269, 0.104 and 0.112....so the red is quite a bit higher than the other two.

So this puts the three histogram peaks on top of each other.

At this point I can just apply STF (three channels locked together) and I see the small green stars....OR I can go through a full BackgroundNeutralisation and ColorCalibration process...and I still see the small green stars.

I have tried varying just about every setting in DSS...always get the small green stars....and anyway, I get the same result if I look at just a single RAW sub....which sort of rules out DSS.

Any ideas as to what I am doing wrong?

Cheers
         Simon

[Juan Conejero]

Hi Simon,

This is a color calibration problem. Could you upload a significant crop of the linear image?

[Simon Hicks]

Hi Juan,

The orignal autosave file is 116MB, and my upload site has a limit of 50MB per file. So I cropped to about the central thrid of the image. This is the untouched file as saved by DSS with no adjustments.

Cheers
         Simon

http://cid-7754931c69f7958d.photos.live.com/self.aspx/Astropics/Autosavecrop.tif

[Enzo De Bernardini]

I always had some problems using DSS with DSLR RAW images, and they look exactly like your example. Try to convert RAW to TIFF and then use DSS or, even better, do it in PI. Perhaps a few more steps, but the results are ideal.

Regards,

Enzo.

[Simon Hicks]

Hi Enzo,

That's interesting that you've had the same problem. I will certainly try stacking in PI tonight and report back. But even my subs show the same small green stars and they have not been anywhere near DSS. 

Cheers
         Simon

[Simon Hicks]

Converting to tif and stacking, or stacking in PI makes no difference. And the small green stars are there in the individual subs anyway.

So its either some wierd non-linearity in the sensor or I am doing something wrong with the colour calibration? 

[Simon Hicks]

Juan,

Did you manage to download the upload? Just wondering if there was any feedback?

Cheers
         Simon

[Juan Conejero]

Hi Simon,

Yes, I donwloaded the image and gave it a try. I get the same results: all small stars are basically green.

This can be 'fixed' with SCNR, but that's just a dirty trick solution, not a real solution to the actual problem. I have no idea what it can be, but the fact is that your smaller stars are all green... somebody has a clue? 

[Simon Hicks]

Hi Juan,

Many thanks for looking at this. In a way I am relieved...I thought I was doing something really silly!

I have tried applying SCNR but it really doesn't seem to do much at all to them. I have tried  mountains of different things and the best I could do was just select all the very small stars with a mask and then apply ColourSaturation at -10 to try and get rid of all colour from them. Sledgehammer / nut comes to mind. Yes...a bit of a shame to have to do that.

Could there be something wierd about the Canon 1000D? Could it have some sort of pedestal on the green channel that is not on the red or blue channels?  Or maybe the Red and Blue channels have a negative pedestal applied?

If anyone has any ideas then I'm all ears!

[Carlos Milovic]

Could you upload a raw frame?

[Simon Hicks]

Here you go.....

http://cid-7754931c69f7958d.photos.live.com/self.aspx/Astropics/IMG%5E_0005.CR2

[Carlos Milovic]

Hi Simon
It seems to me that the red channel is out of focus. This may explain why the cores look green. Also the blue channel is a bit "spreader" than the green one, but a good color calibration may deal with that.
The workaround I propose: deconvolute the R channel, trying to make the stars tighter. Then, fix the green cast with SCNR (still in the linear image, color balanced), preserving the luminance (do not use SCNR's option. Do that yourself).

[Simon Hicks]

Hi Carlos,

Thanks for your help Carlos, that is greatly appreciated.

I took a small crop, and extracted the three RG and B inages...and yes, the small red stars are much bigger than the green ones...well spotted!

I tried solving the problem by bluring the green stars slightly...which is always easier than tightenting the red stars....and then recombined. I had a little success with this, but not as good as I would have hoped. I'll try the deconvolution route and see how I do.

The question I'm left with is...how do I get good RGB data with this lens? Its a Pentax 67 150mm lens....extremely flat field and I like it a lot. Its extremely good with narrowband imaging. I usually stop it down from F2.8 to F4 when narrowband imaging. I wonder if I should be stopping it down even further to try and get the full spectrum RGB to be in focus? Maybe that is the best bet with these images....i.e. just try reshooting them with a higher F#?

Cheers
         Simon

[Carlos Milovic]

Test the lens performance at various f stops. Then decide what is best for you. At the end, it is always a compromise between faster exposures or less aberrations.
Another solution, of course, would be to take another shot focusing the red channel... but since focusing common lens is a p**n ** *** **s, I would not go down that road