Star Halos in RGB

[brew]

I am encountering a problem which I don't know how to solve

I have collected a series of RGB frames, 15 minutes in each filter. The hardware is Tak Sky90 refractor, SBig ST2000XM camera. System is refocused on filter changes, as well as every 4th frame. Calibration was done in MaximDL when the images were collected. Bias, Dark (same temperature) and Flat frames applied.

Alignment and integration were performed in PI. Images were drizzled, and combined with DrizzleIntegration. Works Great!

I then cropped, and performed DBE on each frame using the same model.

OK, finally the problem:

I do LRGBCombination on the frames (no Lum), then do BackgroundEqualization using a small preview on the background. This works appropriately (I think) BUT many of the stars have blue halos, quite large on the larger stars. I have attached a small preview of the image showing these halos.

I tried using MorphologicalTransformation Erosion on the original Blue image, using a starmask to only affect the stars. I found that the Amount of the MorphTrans caused some stars to be cleaned up; basically the larger stars never got right, but middle stars looked better while smaller stars became orange (too little blue). A larger Amount (0.9) caused more smaller orange stars, a smaller amount (0.5) only affected the smallest stars. The second attachment shows the result of 0.7 Amount. You can see that the larger stars seem about the same, but smaller stars have lost their blue and now have dark halos around them.

So, the question is how do I "fix" this? I see that I might be able to use StarHaloReducer, but would need to apply it to quite a number of stars:(

The fact that some of the stars have small red halos makes me wonder if maybe this is real, like the stars in the Pleiades. The target is the Propeller Nebula, which has a lot of nebulosity in the area.

[Alejandro Tombolini]

Hi, it is probably that a star mask was active when you applied Background Neutralization. If you have saved the project, check it or going throughout History Explorer, you will find where the problem originates.
Saludos, Alejandro.

[brew]

Hmmm, I do not see a mask when I look in History Explorer.

I did use a mask when I did MorphologicalTransformation. The mask is still active when I do the LRGBCombin; I guess I expect that the mask doesn't affect the combination.

[Jason Tackett]

One thing that comes to mind…Recently there was a bug fixed with DrizzleIntegration where stars are expanded when there is an image in the stack that is rotated by 180 degrees (link below). Perhaps this may be affecting the integrated image in the blue channel? The bug has been fixed with the update on (or around) July 8th.

http://pixinsight.com/forum/index.php?topic=7271.0

Jason

[brew]

All of these images are oriented the same way. ACP/Maxim flips images after a GEM meridian flip so everything is consistent.

I did find that if I adjusted the Blue coefficient in LRGBCombination to be 0.6 that things are much better. There are still a number of larger stars with halos, but overall the image is much better. This suggests I should be doing the eXcalibrator type of subframe balancing. I thought this wasn't necessary in PixInsight because ColorCalibration ends up doing a similar thing.

I tried eXcalibrator to see what coefficients it generates (maybe a 0.6 for blue?). Of course, it won't run for me. It indicates that the SDSS databases don't cover my FOV. I don't know what this means - is my FOV too large or too small, or do they skip areas of the sky?  When I try Nomad, the program encounters an integer overflow problem in processing the background:( I have saved the images as 32 bit integer FITS; I guess now I get to try various combinations of FITS format, crops of the images, etc. Anyway, I will move my eXcalibrator difficulties to some other forum.

[Juan Conejero]

This problem has nothing to do with color balancing. If the halos exist in the blue and red channels of your data set you can't fix them with color calibration tools; you can only try to dissimulate the problem with selective processing via star masks. I can't say anything more useful without inspecting the raw data.

[Juan Conejero]

The mask is still active when I do the LRGBCombin; I guess I expect that the mask doesn't affect the combination.

That depends on how you apply the LRGBCombination process. If you execute it in the global context (F6 = blue round icon), then an active mask in any of the source images does not have any consequence. At the contrary, if you execute it locally on the RGB image for example, then an active mask in the target image will be applied. As Alejandro has pointed out, this could be the cause of the problem.

Another frequent error with LRGBCombination is to use it on linear data. The LRGB combination process requires nonlinear (stretched) images. Furthermore, the nonlinear RGB and lightness images should be made compatible before combining them. I strongly recommend you watch this video that describes the whole procedure.

[brew]

1) I happened to use the global circle to combine the images, so I think I am OK there. In any event, I re-did the combination without the mask being present and got the same result.

2) I am confused on the second point,

Another frequent error with LRGBCombination is to use it on linear data. The LRGB combination process requires nonlinear (stretched) images. Furthermore, the nonlinear RGB and lightness images should be made compatible before combining them. I strongly recommend you watch this video that describes the whole procedure.

Yes, I am combining RGB (no L) in linear mode (no stretches applied yet). The statement indicates I should combine stretched images; however, the video shows combining linear data. The statement and the video seem incompatible - am I missing something?

I will attach jpegs of my R,G,B masters, maybe they can help. The originals are too big (30Mb each) and I don't have an easy way to provide access..

[brew]

And here is the red

[Alejandro Tombolini]

Hi,

Yes, I am combining RGB (no L) in linear mode (no stretches applied yet). The statement indicates I should combine stretched images; however, the video shows combining linear data. The statement and the video seem incompatible - am I missing something?

The video shows how to convine a linear RGB image using LRGBCombination Tool. With that you will have your RGB linear image.
After that, to obtain your LRGB image using LRGBCombination tool, you need to use a non linear data (RGB and Lightness).

The originals are too big (30Mb each) and I don't have an easy way to provide access..

See this article about Sharing Images in PixInsight Forum.

Saludos, Alejandro.

[brew]

OK, I see the distinction -
RGB with LRGBCombination should use linear R,G,B. This is what I was doing, so I am OK there.
LRGB using LRGBCombination should use stretched images. That is what I do once I get to that point, but I never realized the distinction.

Hmmm, I tried to log into the Endor / Pydio site, but had no luck. I have sent a Contact message to try and understand which user name/Id and which password I am supposed to use. Or, maybe I am supposed to get a third login for the Endor process? A bit unclear.

[brew]

OK, I now have my Endor account. I am uploading the three file R,G,B; it looks like it will take awhile:)

I think the link to the folder is http://endor.uv.es/files/data/public/14085f.php, if I understand this.

[Juan Conejero]

This is the RGB combined image with your data:

Click for full size image

And here are the individual RGB components for comparison:

Click for full size image

Click for full size image

Click for full size image

The blue channel has bloated flat stars and lacks all dim stars, compared to the red and green channels. This is a focusing problem in my opinion.