StarTools 1.9 preview

[cytan299]

Hi Ivo,
I just downloaded 1.9 alpha on my Mac (experienced a crash already in the Sharp module, hopefully, it's a one-off. If it crashes again, I'll file a report). Anyhow, I have a question:
In the Compose module, there is a new option: Spectrum and Filters.

So with this new option, when I use my L-extreme filter, I should specify
Spectrum and Filters->DUO/TRI/QUAD band
and
Luminance, Color -> L+Synthetic L from R(2xG)B, R(GB)GB) (Bi-Color from OSC/DSLR)

Is the above right?

Also, what does the new option do? Does it help compensate colours from Light Pollution Filters if used, i.e. Spectrum and Filters-> CLS or UHC filter?

cytan

P.S. Thanks for the new version before X'mas.

[admin]

In the Compose module, there is a new option: Spectrum and Filters.

So with this new option, when I use my L-extreme filter, I should specify
Spectrum and Filters->DUO/TRI/QUAD band
and
Luminance, Color -> L+Synthetic L from R(2xG)B, R(GB)GB) (Bi-Color from OSC/DSLR)

Is the above right?

That's correct.

Also, what does the new option do? Does it help compensate colours from Light Pollution Filters if used, i.e. Spectrum and Filters-> CLS or UHC filter.

It makes StarTools aware of what type of dataset this is, which has an effect on how the UI appears in certain modules (the Color module is a good example; in the UI it will display "H-alpha" and "O-III" instead of "Red", "Blue" and "Green"). It furthermore activates presets that are more suitable for the subject and spectrum.

[decay]

Saying that, I am actually looking for feedback on the SVDecon workflow/UI. The automatic determination of sample size and automatic apodization mask generation in particular (any failure cases). Note also that you should now be able to select stars that have overlapping sampling areas (if need be).

I nearly missed your post

I lost some time trying to set up a possibility to do a side-by-side comparison as ST doesn't allow to run two instances. But what I already can report: No crashes or failure cases so far. That looks all fine I'm always trying to click on the 'Sampling' button right after module start (as I'm used to do from 1.8), but then the mask generation screen comes up and surprises me .

There seems to be a problem with the 'starfishies' selection which pertains in the following up modules (e.g. Color module). I was not able to get rid of it and generate a 'normal' star mask for colour sampling.

I will later report back regarding SVDecon processing itself.

Best regards, Dietmar.

[Mike in Rancho]

Hmm, might be a bit hard to keep it to substantive commentary. Not sure where it crosses the line into UI? Other than the obvious stuff that just "isn't ready yet."

Dietmar yes, that issue on the Mask settings was also in 1.8 and was noted as a bug to be fixed. For me the easy way out now is to just hit the FatStars preset then immediately the Stars preset, and it should unset starfishies and be back to normal. As you say, the most common location for that to rear its head is wanting round stars for Color sampling immediately after SVD. Ivo explained it in another thread, but I tried turning off one of the starfish sliders (forget the name, upper right control) and that wasn't sufficient.

SVD's intelligent starfishy generation and sample size has been pretty good so far, maybe a bit hit or miss on oddball datasets. Say with a lot of small stars. I'm not sure what it is looking for, but it seemed it could be biased a little too small, so it never white-circled a number of stars I thought would be good candidates, and I just had a ton of outlined junk stars. I was able to find some good ones throughout the image, but I didn't have good spatial coverage.

I still need to run 1.8 and 1.9 side-by-side (different computers, no virtual box ) on varying datasets to come up with better commentary.

I've also been working with JWST data lately (more on that later), which as you can imagine can be a bit unusual, especially as to stars.

Here's a weird one, substantive? Sharp module crashes at mask generation if used after or outside of tracking. Or so it seems. So - if you finished denoise and decided hey I want to add a tiny bit of sharpening after the fact (without undoing or going back), or, loading already stretched data at the beginning for open file.

[hixx]

I'm not sure what it is looking for, but it seemed it could be biased a little too small, so it never white-circled a number of stars I thought would be good candidates, and I just had a ton of outlined junk stars.

Hi, I have also noticed that, Mike. Some candidates seemed pretty green to me, yet got not selected for whatever reason, But a lot of faint stars get selected. Is there a way for ST to further narrow down by using lower / upper tresholds " not dimmer as.." &"use green/yellow only", So it shows fewer, but better candidates?

I'd have a suggestion regarding the color coding in sampling view though:
While the red/yellow/green scheme might be familiar to anyony thanks to traffic lights, it is hard to distinguish green from yellow.
Probably helpful would be pulling yellow more towards amber and pull green more into teal. Red may be replaced by purple/lilac, to give a stronger "DON'T!" signal, even in small cores.....

Clear Skies, everyone!
Jochen

[Mike in Rancho]

Hi, I have also noticed that, Mike. Some candidates seemed pretty green to me, yet got not selected for whatever reason, But a lot of faint stars get selected. Is there a way for ST to further narrow down by using lower / upper tresholds " not dimmer as.." &"use green/yellow only", So it shows fewer, but better candidates?

I'd have a suggestion regarding the color coding in sampling view though:
While the red/yellow/green scheme might be familiar to anyony thanks to traffic lights, it is hard to distinguish green from yellow.
Probably helpful would be pulling yellow more towards amber and pull green more into teal. Red may be replaced by purple/lilac, to give a stronger "DON'T!" signal, even in small cores.....

Clear Skies, everyone!
Jochen

NO should be....lilac?....steada red?

Though I get your point. Sometimes it's hard for me to see if there's a tiny red dot inside a yellow core. But that's more from it being tiny, so I have to put my face up to the screen.

SVD 1.9 still seems to do a fine job, even choosing seemingly poor stars that are red or orange blobs devoid of green or even yellow. But yes sometimes there's just nothing really usable in large areas, and while there may be outlines, there don't seem to be recognizable stars within them.

Honestly though I'm still a neophyte at starfishy-SVD, so I'm not always terribly sure what I should be looking for. Based on the 1.9 description we can now overlap - the blue sample squares? - and SVD will still figure out what goes with what?

But I'm curious about the white outlines in general. I know from before that the stellar profile area was what falls both within the outline and within the blue sample box. But -- sometimes the white outlines are a very scrambled jigsaw puzzle, such that it is hard to tell what is IN and what is OUT, in order to then apply blue boxing and know what it is thinking. I haven't seen any poor results yet I think, so I figure SVD is smarter than I am.

[decay]

that issue on the Mask settings was also in 1.8 and was noted as a bug to be fixed

Oh, OK Mike, I must have missed that in the discussion of the starfishies. And thanks for the hints to workaround. That's important to know since using starfishies is now mandatory.

Best regards, Dietmar.

[hixx]

such that it is hard to tell what is IN and what is OUT, in order to then apply blue boxing

Hi Mike,
the good thing about 1.9 is you actually don't have to worry about IN/ OUT or any sample size box, as this is what the jigsaw puzzle does. If I got the release notes correctly, the PSF "mask" algorithm is drawing shapes of candidates on its own, without user intervention.

As per what candidates to select, I simply use anything that looks like a star im each corner, borders and center which is not awfully red (and no, I don't insist on purple, I'm just looking for a distinctive color scheme)
Cheers,
Jochen

[Mike in Rancho]

Hi Mike,
the good thing about 1.9 is you actually don't have to worry about IN/ OUT or any sample size box, as this is what the jigsaw puzzle does. If I got the release notes correctly, the PSF "mask" algorithm is drawing shapes of candidates on its own, without user intervention.

As per what candidates to select, I simply use anything that looks like a star im each corner, borders and center which is not awfully red (and no, I don't insist on purple, I'm just looking for a distinctive color scheme)
Cheers,
Jochen

Thanks Jochen. Well, I'm still chewing on what exactly "SV Decon no longer requires separation of samples" from the changelog means. Granted were jumping the gun a little bit on this preview, and documentation (and updated User Notes!) will ensue at the appropriate time.

But while the sample box size and apod mask are off the table, I think we still have to know what the box/outline regime means (especially in weird jigsaw puzzle situations) because picking the right stars is still up to us. And therefore not picking those that are kind of funky.

But it seems to be working pretty well no matter what I do...

[Startrek]

Hi Ivo,
I just read through all the changes /improvements for 1.9 and I must say it’s very comprehensive so can’t wait until fully tested and ready
SV Decon in 1.8 is already excellent so any improvement will be well received
I’m glad you brought back DSO core in HDR , I really liked that function in 1.7
All the extras in Compose and Color will be great too

Many thanks and looking forward to this version
Clear Skies
Martin