Color misalignment - can I do something about this in PI?

[magnusl]

Hi!

I have an issue with a slight color misalignment, most visible in the top left corner of the attached image. It is DSLR images fully processed in PI, 87 subs. Is there for instance a way to split the RGB channels and then realign them, that would work here?

Full xisf-image is found here:

https://www.dropbox.com/sh/fs744ujp8oed79d/AACfnztnt-E98N8km8Tv2dUBa?dl=0

Magnus

[magnusl]

Here is the top left corner with most color misalignment.

Magnus

[Nocturnal]

Is that really misalignment or caused by diffraction? In any case, you can split your images into the three channels and use the alignment tools to align two onto 1 and then recombine them. I don't think there is a process that automates it. I'm afraid if the colors you see are caused by 'flares' of color rather than misalignment of the stars there isn't much that re-allignment will do. Give it a try.

[magnusl]

Hi!

Well, I have no idea what causes it - my guess would be diffraction (I suspect my image plane is actually somewhat tilted).

Can you say a little more about how I could combine them? I follow you in splitting in R, G and B channels. But then what? Which tool to use? I've heard something about Affine transformation.... but it doesn't tell me much.

Magnus

[Nocturnal]

Hi,

I think it makes sense to look to the StarAlignment tool, right? I would see if 'distortion correction' helps in your case.

[bulrichl]

Hi Magnus and Sander,

this doesn't look like diffraction to me but rather like transverse (or lateral) chromatic aberration https://en.wikipedia.org/wiki/Chromatic_aberration , an image defect caused by optical lenses. Pure reflectors don't have this sort of image defect. Which telescope did you use? Did you use a coma corrector, a flattener or (improbable) a reducer? Maybe a low quality optical element with lenses could be replaced by a better quality one or it could be a matter of right adjustment (e. g. the distance between a flattener and the sensor plane is very critical).

I think you can moderate it by image processing as Sander described, but better (if possible) is to elimiate the cause.

Bernd

[Nocturnal]

Hi Bernd,

Thanks for the correction. I had the same effect in mind but used the wrong name. I figured it was caused by the different diffraction of different wavelengths in the glass elements. It would also be possible to be affected by atmospheric diffraction but that depends on the altitude of the target. I agree with you that eliminating the cause the best course of action but we are mere mortals and sometimes we have to resort to fixing after the fact :-)

I think pure reflective and properly corrected scopes are rare. Generally you need lenses to correct for aberations introduced by the mirrors. Of course the quality of the lenses matters a great deal.

[magnusl]

Hi!

It's a C8 with a 6.3 reducer. I guess the effect has to do with the reducer and a slightly tilted image plane (that is my hypothesis). It's a somwhat longer project to experiment with correcting it, but any suggestions on how to do that or how to explore it is welcome.

As for the PI process with Star Alignment: I'm at loss here. I split the image into RGB channels. THen I enter them into the StarAlignment, click on Distortion correction, and run Global. THat gives me 2 new images (the one used for reference produces an error - I am using the views). Then what? Image integration on these produce a greyscale image. CHannel combination works, but I cannot see any difference. Sorry for being lost here.... I never worked with separate R G and B channels before.

Magnus

[Nocturnal]

The fact that they are RGB channels is immaterial. You simply need to learn how to align one image to another. Look for a tutorial somewhere here on the site or elsewhere in internet land. You pick one of the 3 channels as your reference frame, align the other two and then combine the reference frame with the 2 aligned ones back into a color image. You can also use your stacking software to do this like DeepSkyStacker.

[pfile]

channelcombination was the right thing to do (on the 2 newly registered images + the reference image). it's possible that distortion correction did not do too much - there is a setting for what registration model to use; you might try setting it to 2D splines and repeat the process to see if anything improved.

rob

[msmythers]

This is not a solution but something you can do if you need to. Just use a halo mask and reduce the color saturation of the halos. My preference would be to do something like this before stretching the image. You can use whatever method you want to generate the halo mask also, we have many, many ways of doing that in PI.

Again this is just to deal with images that already have the issue. The best solution is find the optical problem and correct it.



MIke

[bulrichl]

Hi Bernd,

Thanks for the correction. I had the same effect in mind but used the wrong name. I figured it was caused by the different diffraction of different wavelengths in the glass elements. It would also be possible to be affected by atmospheric diffraction but that depends on the altitude of the target. I agree with you that eliminating the cause the best course of action but we are mere mortals and sometimes we have to resort to fixing after the fact :-)

yes, I was also wrong and you are right, I had only "atmospheric refraction" (not diffraction) in mind when I read your text.


As Mike meanwhile wrote: I also think that the splitting into RGB channels, aligning and combining has to be applied at linear stage.

Bernd

[Niall Saunders]

Hi Magnus,

Another way to try and 'align' the three channels (once extracted from your RGB image) is to use

Script | ImageAnalysis | ImageSolver (on your RGB image, or the L-channel extract thereof) to determine the optical characteristics of your OTA

Script | Render | CatalogStarGenerator to create a 'synthetic' reference frame against which you will then align the R, G and B planes

Process | DynamicAlignment on each of the three (R, G and B) channels, aligning these against the synthetic image you have acreated above

You can then recombine the R, G and B channels to see if you have improved things in any way.

[Nocturnal]

Hi Bernd,

You can split/align/combine at any stage. I do think it's worthwhile to do it early on, right after stacking but only because you then work with the final image data for the rest of post processing. DeepSkyStacker has an optional channel align step at the end of stacking and AutoStakkert does as well.