just wanted to share this:
-
- ezgif-4-0fcf480702.gif (42.48 KiB) Viewed 16697 times
Very cool!
Best regards, Dietmar.
SV Decon Centroid Tracking
SV Decon Centroid Tracking
Hi all,
just wanted to share this:
I've no idea, what's causing this off-centred shadows, but the 'PSF Resampling' 'Centroid Tracking' Option does an awesome job fixing this.
Very cool!
Best regards, Dietmar.
just wanted to share this:
I've no idea, what's causing this off-centred shadows, but the 'PSF Resampling' 'Centroid Tracking' Option does an awesome job fixing this.
Very cool!
Best regards, Dietmar.
Re: SV Decon Centroid Tracking
The off-centered ringing can sometimes be caused by a dud sample somewhere. However, it can also (more commonly) be caused by pushing a "naive" decon too hard.
In both cases it is caused by the PSF progressively no longer being representative of the true correction that needs to be made every iteration.
The centroid tracking feature tries to track the samples (stars) you set as they "move" during each iteration, after which they are resampled for each iteration. The process therefore becomes self-correcting; any errors that creep in will creep into the corrective model and will therefore be reversed during the next iteration.
It is recommended to always switch this feature on before keeping the result, as it gives almost always better results with less ringing. It takes up a fair bit more processing power, so switching it on later in the module can keep processing times down for when you are still setting samples and evaluating the results from other parameters/settings.
Sidenote; I haven't really seen this feature/capability described anywhere else, so it may be unique to ST...
In both cases it is caused by the PSF progressively no longer being representative of the true correction that needs to be made every iteration.
The centroid tracking feature tries to track the samples (stars) you set as they "move" during each iteration, after which they are resampled for each iteration. The process therefore becomes self-correcting; any errors that creep in will creep into the corrective model and will therefore be reversed during the next iteration.
It is recommended to always switch this feature on before keeping the result, as it gives almost always better results with less ringing. It takes up a fair bit more processing power, so switching it on later in the module can keep processing times down for when you are still setting samples and evaluating the results from other parameters/settings.
Sidenote; I haven't really seen this feature/capability described anywhere else, so it may be unique to ST...
Ivo Jager
StarTools creator and astronomy enthusiast
StarTools creator and astronomy enthusiast
Re: SV Decon Centroid Tracking
Hi Ivo, thanks for the detailed explanation and advice. From now on I will always switch that on for the final decon processing.
I will try to check if this is due to a dud sample. Of course, data is weak (once again for me
) and maybe that's a reason for having problems with star sample selection. There were only a few usable starfishies and most of them only on one side of the image ...
Best regards, Dietmar.
Eye-catching for me is that the orientation of the ringing artefacts is the same for all stars. I would assume that creeping in errors caused by pushing decon too hard would have a random origin and therefore a random orientation?
I will try to check if this is due to a dud sample. Of course, data is weak (once again for me
) and maybe that's a reason for having problems with star sample selection. There were only a few usable starfishies and most of them only on one side of the image ...Best regards, Dietmar.
-
Mike in Rancho
- Posts: 1166
- Joined: Sun Jun 20, 2021 10:05 pm
- Location: Alta Loma, CA
Re: SV Decon Centroid Tracking
Centroid tracking is for sure a great first step in deringing, though with the extra time it does become a "when to turn it on" issue. Depends how big the resolution is I imagine.
Some of the funky shadows may smooth out as 1.9 becomes more refined in further alphas? Things still don't always pan out right for a nice sampling field, depending on resolution, and what the auto-chosen sharfishy size and blue box size is (both variable?).
Having some similar issues last week, I was wondering if what we learned from 1.8 carries over - as far as what gets sampled for PSF is what is bounded by both blue box and white outline? Because sometimes when the sizes don't come out right, some larger white outlines can end up with a seemingly tiny-seeeming blue box within it. Could that "cut off" the outer edges of a star's PSF from the SVD modeling?
Some of the funky shadows may smooth out as 1.9 becomes more refined in further alphas? Things still don't always pan out right for a nice sampling field, depending on resolution, and what the auto-chosen sharfishy size and blue box size is (both variable?).
Having some similar issues last week, I was wondering if what we learned from 1.8 carries over - as far as what gets sampled for PSF is what is bounded by both blue box and white outline? Because sometimes when the sizes don't come out right, some larger white outlines can end up with a seemingly tiny-seeeming blue box within it. Could that "cut off" the outer edges of a star's PSF from the SVD modeling?