SV Decon again

[Startrek]

I’ve processed a few images now using SV Decon and noticed one obviously limitation with this module compared to Decon V1.7
If I Bin my images below 71% I find that some medium to larger stars develop artifacts ( a few dark pixels in the core ) and lose their round shape due to some darker pixels around the edges ( end up jagged looking ) not to be confused with ringing
As far as reducing atmospheric blur , SV Decon works beautifully, but my stars suffer particularly if I want to Bin to improve SNR at the expense of some loss in resolution
My image scale is 0.86 arc sec per pixel ( do oversampled) and in most cases I Bin my images between 50% to 35% to improve SNR but I’ve found this has a detrimental affect on SV Decon
I’ve mentioned this before and others have advised about it but my Stars look like white discs after SV Decon and this is suppose be scientifically correct but it does make the image look a little strange as not all stars are white discs only the medium to larger stars ( something to do with PSF )
With Decon V1.7 Binning was never an issue in my images , reduction in atmospheric blur worked well and stars remained pretty much the same, unless I pushed the pixels up too much.
When I get time I might post some images of my Stars both with SV Decon 1.8 and Decon V1.7


Clear Skies
Martin

[Mike in Rancho]

Yeah, the tight star spots I don't mind, just like thin and sharp diffraction spikes. I don't have the need for soft fuzzy stars.

But, as you say, it's when SVD picks up on something(s) and turns the star cores not into tight spots, but causes either white-ringing (different from the shadow ringing that can be de-ringed) or odd shapes - that look like blocky video grame graphics from the 80's.

Sometimes even turning all the parameters way down doesn't help, and by that time the clarity gained on other objects is lost to. At that point I either note where the defects are for later repair (heal, or layer blurring, whatever it takes), or I'll cancel out and re-enter to just do synthetic, which usually does not cause those issues. Though on some data I even have to dial back the synthetic radius.

And then, other datasets just work flawlessly in SVD and could conceivably even be dialed up from the defaults.

So...probably the data? I haven't necessarily figured out a good bin/not-bin strategy. It also seems to me that SVD is a bit more amenable to refractor data than reflector data.

[hixx]

Hi Martin,
some remarks here:
1) When binning, always make sure You remain in oversampled territory if You intend to Decon. No Decon will be able to bring back information lost in undersampling.
As a rule of thumb You may use the pixel resolution Your optical train would have on the camera. I.e. if the resolution of Your scope was 1 arcsec, image scale of Your Your combo should be close to 1 arcsec / pixel as possible, for SVDecon maybe slightly below (0.8 arcsec/px. This varies strongly with the telescope / Cam combo used, of course. Here`s a formula:
Image scale (arcseconds per pixel) = 206 * camera pixel size (microns) / Focal length (mm).
Should Your image scale be above this limit, Decon is unable to reveal more detail. Please also see the discussion " Decon vs binning " in the manual.

2) the "white ringing" is actual not an artifact but a property of the data (caused by ANY light wavefront passing through a circular hole while imaging ("Airy Disk"!) It is just that SVDecon is able to remove all blur and bring the resolution down to this physical limit. Whether one likes this look or not is a matter of taste, of course.

3) Slight artificial blur may be (re-)applied using a Gauss Lowpass in Layer module or simply using the Atmospheric models of SVDecon (no sample star selected!) to emulate the 1.7 Decon look. Both blur and inferior decon would hide the Airy Disk.

clear skies,
Jochen

[Mike in Rancho]

2) the "white ringing" is actual not an artifact but a property of the data (caused by ANY light wavefront passing through a circular hole while imaging ("Airy Disk"!) It is just that SVDecon is able to remove all blur and bring the resolution down to this physical limit. Whether one likes this look or not is a matter of taste, of course.

clear skies,
Jochen

Hi Jochen. Is that really the case? I am not a scientist, but my impression has been that the Airy Disk actually is a solid disk (spot), and if you could get it perfect, would be the size of the detail resolution that can be achieved by the telescope. No?

Surrounding that Airy Disk/spot, would be a series of fading concentric rings. The ability to see or resolve those rings, I thought, pretty much required high magnification. And imaging them, is generally a "lucky imaging" thing meaning the same way you would do planetary. At least for mundane amateur telescopes, anyway.

It doesn't happen often, but when it does (slightly dimmer core surrounded by bright white ring), I've assumed it to be an SVD artifact, possibly due to core saturation during acquisition.

Me wrong?

[Startrek]

Thanks All,
I have a better understanding of SV Decon and I know how to use it now but …..
I guess the main point I wanted to make is that out of all of the modules in ST , SV Decon has its limitations in that your image data must conform to specific criteria or else you may not be happy with results , in particular the resultant stars appearance
SV Decon in reducing atmospheric blur is excellent no matter what type of data but it’s the Stars appearance that to me is the main factor.
Whether it’s scientifically correct or this is how it should look , I still maintain that stars look a bit artificial like saturated white discs and from a non ST user they may suggest you have just been too heavy handed in your processing.
The fact that I Bin my data is one limitation, if I don’t Bin, my image size is huge and noisy, but SV Decon works in my favour with better results on Star appearance. If I do Bin , my image size is more manageable, less noisy but my Stars appearance is terrible
By the way my image scale at my observing locations are

Bortle 8 City suburban 0.86 arc sec per pixel
Bortle 3 Dark site 0.62 arc sec per pixel

So I’m definitely oversampled

Moving forward I’m just going to have to use SV Decon Synthetic option so my stars look half normal
It’s a shame because the atmospheric de blur works so well on the Sampled method , it’s just the look of those Stars and it gets worse with more Binning

All the feedback has been greatly appreciated
Looking back through posts this year , I’m glad I’m not the only one who has had some sort of issue with SV Decon

Clear Skies
Martin

[admin]

Hi Martin,

Just so I'm not missing something, would you be able to clearly demonstrate the issue for me with two identically processed images?
Perhaps, just a tiny crop of a dataset, processed identically in 1.7 and 1.8, showing the issue. If there may be anything amiss, would you be able to put that tiny little dataset/crop aside, so you could perhaps share it with me if necessary?

Thank you!

[Startrek]

Hi Martin,

Just so I'm not missing something, would you be able to clearly demonstrate the issue for me with two identically processed images?
Perhaps, just a tiny crop of a dataset, processed identically in 1.7 and 1.8, showing the issue. If there may be anything amiss, would you be able to put that tiny little dataset/crop aside, so you could perhaps share it with me if necessary?

Thank you!

Hi Ivo
Yes no problem
When I get some time I’ll post some comparison cropped areas of 1.7 Decon and 1.8 SV Decon labelled using my latest M8 Lagoon image from a week a go
Give me a day or two
I’ll post in Gallery

Thanks
Martin

[admin]

No stress, much appreciated!