Automatic Coma correction

[Guest]

Since the coma effect is completely predictable on newtonian scopes, I wondered if it was possible to correct for it in the imaging software. Or has someone already done that?

Cheers

Steve

 

[Adam J]

20 minutes ago, woodblock said:

Since the coma effect is completely predictable on newtonian scopes, I wondered if it was possible to correct for it in the imaging software. Or has someone already done that?

Cheers

Steve

 

It might be possible to train an AI to make the correction as is the case with Starnet++ but to be honest that would not solve the issue of distortion to the background DSO. 

[vlaiv]

It is indeed possible to do it but I don't think that anyone has done it.

It would involve deconvolution (already exists implemented) with deconvolution kernel depending on position on image.

Although in principle one knows level of coma in newtonian scope with parabolic primary, in practice things are not as easy. Level of coma depends on collimation of the scope and position of the sensor. If sensor is slightly shifted with respect to optical axis (not tilt, but rather shift - so that optical axis does not go thru exact center of sensor), aberrations will not be symmetric with respect to sensor.

One can account for that by examining stars in the image and determining true optical axis / sensor intersection. One can also generate coma blur PSF for certain distance from optical axis (for perfectly collimated scope), so yes it can be done.

Downside is that it is inherently probabilistic process because data that you have suffers from noise - you are trying to guess rather then precisely calculate because you don't have exact numbers to start with, but rather values polluted by noise. Another difficulty would be that it is better to do it on stack of data rather than single sub (better SNR) but stack of subs will have different levels of coma if you dither or otherwise have less than perfect alignment - like slow field rotation / drift / whatever, because it changes distance of pixels from optical axis.

Result will be corrected image but lower SNR - very similar to what you get when sharpening - sharper but noiser image.

That applies to any sort of optical aberration - as long as you have proper mathematical description of it (like astigmatism depending on distance from optical axis, or even field curvature) it can be done with deconvolution - at expense of SNR. It is much easier to correct it with optical elements (less costly) whether that is coma corrector, field flattener or whatever ...

Just to add - above method will deal both with star coma blur but also extended features coma blur because it will operate on precise mathematical definition of coma blur rather than approximation of neural networks such as StarNet++

[bottletopburly]

Startools can heal starts making oval stars round again @jager945