Stellar Color calibration - Not enough matches found

[roelb]

Hello Robin.
For the first time I had the opportunity to test the new Stellar Color Calibration feature.
I used a ZWO ASI533MC-Pro camera with following image controls:
Brightness: 10
White balance (R): auto 50
White balance (B): auto 50

PROBLEM 1:
When the feature works all my images turn out too blue.
Then when using the "Auto Colour Balance" in the Histogram, the colors where much better.

PROBLEM 2:
I attached an example where the Stellar Color Calibration didn't resolved a solution despite 39 catalog stars in FOV.
On my mini PC at the scope I didn't had the possibility to take a screenshot, so I used the Folder Camera to simulate the proces.
Relevant files can be downloaded here: https://transfer.it/t/vfeXBG9TiZaK.

Test where done at Bortle 4.

Thanks helping me further.

[admin]

Hi Roel,

the first question is - are you using any sort of filter (particularly of the 'multi band' type) when imaging? At the moment I know that those filters tend to cause problems for the calibration because it will try to reverse the colour cast they apply to the stars in the image. I am waiting for some clear skies to run some real tests to see if I can find a workable solution to that particular issue.

I should note that if you run the colour calibration first and then the old 'auto colour balance' then you are undoing pretty much all of the 'cleve stuff' done by the calibration - the calibration does three things

* Sets the R/G/B vertical sliders for colour balance based on the correlation of star colours
* Auto sets the R/G/B background subtraction (very similar to the 'Auto Adjust' button in the background subtraction tab)
* Sets a fairly strong auto live stack stretch (similar to the histogram auto stretch, but stronger)

Using the old colour balance afterwards resets the first item, and the other two are things that you could do elsewhere in live stacking anyway...

If you aren't using filters and can share one or two of the images where it gives a blue cast, that would be interesting to investigate.


On to the one where it doesn't give a solution...

I think there are a couple of things going on here

* you have a relatively small FOV of about 0.55x0.55 degrees, meaning that there are only a relatively small number of calibration stars in the catalog in that area (39 according to the output). Those are likely to be among the brighter stars in the area

* The stars in the image are fairly bright - it's possible that all of the 39 calibration stars are at least partially saturated in the image, meaning that they are unsuitable for use for calibration - this means that you get zero overlap between the 39 catalog stars and the image stars that are not saturated.

Now, the 39 calibration stars come from the default SharpCap plate solve index, which now contains colour data (for about 93% of the stars in the index). To get more calibration stars, what we need is a 'colour data' version of the high res index (which I have already made, but not released) and some code in SharpCap to actually use that high resolution index when it has colour data (which doesn't exist yet). At least I hope that will help resolve that issue.

Since you have already shared the problem image with me, I will try a quick test later to see if using the high res index does help with that image (and if not, I can investigate other possible causes) - I will report back!

cheers,

Robin

[admin]

Hi Roel,

actually, do you have an unprocessed, stacked, FITS file saved from the live stack that you can send me - what I mean by this is one saved using the 'Save as 16 bit stack' or 'Save as 32 bit stack' in the live stacking save button?

The files you shared are a stretched colour PNG (which I can't run the calibration code on as it is already stretched/processed) and the FITS frame looks like a single raw frame from the camera that will be too noisy to give realistic results.

If you haven't got the saved stack but have the raw frames, you could re-run the stacking in the folder monitor camera to create the 16 bit save.

cheers,

Robin

[roelb]

I don't use any filters, dont't use darks ...
The fits file is a 16 bit saved stacked raw frame.

I hope clouds will disappear this evening and do more tests.

[roelb]

So, I must save the UNPROCCEST 16 bit fits?

[admin]

Hi Roel,

ideally, for this sort of issue, the stack saved with the 'Save as 16 bit stack' option is best - that saves the stack without the stretch/colour balance/enhancements, so that it is the same 'raw but stacked' data that the calibration works on inside the application.

In this case though, I had another look at your single frame FITS file last night and managed to get some results out of it - the colour calibration will not be accurate because of the noise, but that wasn't really the important bit... The important bit was matching up unsaturated stars in the frame with stars from the star catalog that have colour data.

I tried various things with the default star catalog, but I got zero matched stars like you, or sometimes one star - not good enough. I think had a go with a 'with colour data' version of the higher resolution plate solving catalog and that did the trick straight away - I get 74 matched stars and a reasonable looking result from tha calibration calculation.

So, that's good to know - I have updated the code so that from the next version it will be able to use the higher resolution plate solving database as a source of star colours (you will need to re-download that database once the new version comes out, as the current version doesn't have the star colour info in it).

cheers,

Robin

[roelb]

Thanks very much Robin. That is a great job!
Yesterday evening I had the opportunity to do some more tests.
And some of them did have a solution.
But the blue color persists.
I will sent you this afternoon the results together with the screen prints.

[roelb]

Here you can find my latest recordings: https://mega.nz/folder/cadyGZja#M2rZOOyZlNPyZUT1V7sNAg.
CC has found a solution but the blue colour is always prominent.

[admin]

Hi Roel,

thanks for sharing those images - I have had a quick look and I think that the problem is that the calculations for the background subtraction amounts for the R/G/B channels are leaving behind too much blue (not subtracting enough). That gives you the blue effect in the background. I don't think the R/G/B balance adjustments coming out of the colour calibration are the issue - those seem to be giving nice clear curves, which gives confidence in the values that come out.

What I will try to do is dig further to work out why the blue subtraction amount is insufficient. As a workaround for now, I'd suggest going to the Live Stacking 'background' tab after doing the colour calibration and tweaking the blue level downwards (drag the 'Adjusted' position for the blue left) until you get a result that you are happy with.

cheers,

Robin

[roelb]

Thanks a lot Robin.
Until next week I have no chance to observe.
Hope mid next week to do some further tests.
Again thanks for your excellent support.