StarTools 1.9 preview
-
- Posts: 1164
- Joined: Sun Jun 20, 2021 10:05 pm
- Location: Alta Loma, CA
Re: StarTools 1.9 preview
Hmm, well I'm not sure I'm feeling it with the new NB Accent changes.
Could not get a very good result, and it's also nice having the full 0-100 sliders for L and Color to make things "just right."
It's possible I just don't know how to use it yet, or I handled Screen 1 differently. There were a couple little blobs that showed up, more like you'd want for galaxy (this is nebula Ha augmenting), that don't in 1.8, but overall the 1.8 is far better.
The framing here is M78 off to the side, LRGB, with a big curving band of Barnard's loop off to the other side, for Ha enhancement.
Data is poor, just 2.5 total hours last night in awful LP, perhaps poor transparency, a kitchen light I forgot to turn off until halfway through my L filter, X-mas lights everywhere, and my flats may not have worked the best. Good guiding though!
I could provide links if the data may be useful for testing. And hopefully I'll add more integration later this week.
Oh it also seemed to me that the gamma slider was affecting global image gamma rather than just the accenting file.
I did get good SVD starfishies though in both 1.8 and 1.9 - plenty of great sampling candidates to choose from.
Here's what this thing looks like (in ST 1.8) for an idea of what I was trying to do. 1h L, 20m each RGB, and 30m Ha. I didn't save any of my 1.9 attempts for comparison but I could go back and do so. As mentioned, weak data, plenty of noise, but 1.8 did an admirable rescue on it.
Could not get a very good result, and it's also nice having the full 0-100 sliders for L and Color to make things "just right."
It's possible I just don't know how to use it yet, or I handled Screen 1 differently. There were a couple little blobs that showed up, more like you'd want for galaxy (this is nebula Ha augmenting), that don't in 1.8, but overall the 1.8 is far better.
The framing here is M78 off to the side, LRGB, with a big curving band of Barnard's loop off to the other side, for Ha enhancement.
Data is poor, just 2.5 total hours last night in awful LP, perhaps poor transparency, a kitchen light I forgot to turn off until halfway through my L filter, X-mas lights everywhere, and my flats may not have worked the best. Good guiding though!
I could provide links if the data may be useful for testing. And hopefully I'll add more integration later this week.
Oh it also seemed to me that the gamma slider was affecting global image gamma rather than just the accenting file.
I did get good SVD starfishies though in both 1.8 and 1.9 - plenty of great sampling candidates to choose from.
Here's what this thing looks like (in ST 1.8) for an idea of what I was trying to do. 1h L, 20m each RGB, and 30m Ha. I didn't save any of my 1.9 attempts for comparison but I could go back and do so. As mentioned, weak data, plenty of noise, but 1.8 did an admirable rescue on it.
Re: StarTools 1.9 preview
Same here, Decon mask doesn't get the stars as well as 1.8
Here's comparision with same file.
1.8 1.9 Cs
Here's comparision with same file.
1.8 1.9 Cs
Re: StarTools 1.9 preview
Just a really quick note on the "mask generation" while I keep working on improved detection;
The mask generation in 1.9 has evolved a bit from being just an apopdization mask generator.
It is now much closer to a proper PSF selection tool. As such - when it actually latches on to the stars - there are a lot more smarts/statistic at play when the module determines which stars make good PSFs and which don't.
It whittles down the detected star population based on (linear) brightness (not too bright, not too dim, not too much background level), and other measures such as eccentricity and profile gradients.
Only those stars that score well on these metrics make the cut, to become potential sources for a PSF. This means that there are fewer, but higher quality stars to choose from. This should hopefully make it easier and less time-consuming to find good candidates.
One of the important improvements is the "starfishies", which is the result of including more pixels into the PSF. The module has deemed these pixels as likely part of the original point light (you may also see notches or holes where faint companion stars are detected). The more such pixels can be included to reconstruct the PSF, the better, as it will yield a more complete model of the local PSF. 1.8's PSFs masks tended to not include these faint pixels that might still have been part of the stellar profiles.
By being able to keep track of which pixel(s) belong to which star/PSF you can now indeed select stars that are close together where the selection box may overlap without problems. In 1,8 there was no such awareness of which pixel belonged to which star, and thus box overlap was not allowed.
I hope that clears up where 1.9 is heading in terms of PSF selection and how it fits in with new behavior and what you're seeing on-screen.
The mask generation in 1.9 has evolved a bit from being just an apopdization mask generator.
It is now much closer to a proper PSF selection tool. As such - when it actually latches on to the stars - there are a lot more smarts/statistic at play when the module determines which stars make good PSFs and which don't.
It whittles down the detected star population based on (linear) brightness (not too bright, not too dim, not too much background level), and other measures such as eccentricity and profile gradients.
Only those stars that score well on these metrics make the cut, to become potential sources for a PSF. This means that there are fewer, but higher quality stars to choose from. This should hopefully make it easier and less time-consuming to find good candidates.
One of the important improvements is the "starfishies", which is the result of including more pixels into the PSF. The module has deemed these pixels as likely part of the original point light (you may also see notches or holes where faint companion stars are detected). The more such pixels can be included to reconstruct the PSF, the better, as it will yield a more complete model of the local PSF. 1.8's PSFs masks tended to not include these faint pixels that might still have been part of the stellar profiles.
By being able to keep track of which pixel(s) belong to which star/PSF you can now indeed select stars that are close together where the selection box may overlap without problems. In 1,8 there was no such awareness of which pixel belonged to which star, and thus box overlap was not allowed.
I hope that clears up where 1.9 is heading in terms of PSF selection and how it fits in with new behavior and what you're seeing on-screen.
Ivo Jager
StarTools creator and astronomy enthusiast
StarTools creator and astronomy enthusiast
-
- Posts: 1164
- Joined: Sun Jun 20, 2021 10:05 pm
- Location: Alta Loma, CA
Re: StarTools 1.9 preview
Thanks Ivo.
I gather that explains why - in the datasets it is working well for - the population of truly selectable candidate stars (as opposed to small noise starfishies) all seem to be rather similar.
Having mild experience using AltStars while still in 1.8, the notches never really bothered me. But are you saying that the stars with little holes are okay to select also, as SVD knows what it is looking at?
I just had to get my Screen 1 a little more dialed in, and then work the strength slider. I still need to figure out all the available options for Luminance and Color modify and see how that compares with the full sliders in 1.8. The 1.9 gamma still seems to be global however, whereas when I tried the same in 1.8 it seems fully tied to the NB Accenting.
One other tidbit I noticed but haven't fully worked out yet - a late stage OptiDev re-do of global stretch in 1.9 seems to unravel deringing, which presumably was applied in SVD. I don't think I notice that with a late AutoDev re-do (post SVD) in 1.8.
I gather that explains why - in the datasets it is working well for - the population of truly selectable candidate stars (as opposed to small noise starfishies) all seem to be rather similar.
Having mild experience using AltStars while still in 1.8, the notches never really bothered me. But are you saying that the stars with little holes are okay to select also, as SVD knows what it is looking at?
With some more playing around I was able to get some good results with 1.9 nebula NBAccent.Mike in Rancho wrote: ↑Mon Dec 19, 2022 8:09 pm Hmm, well I'm not sure I'm feeling it with the new NB Accent changes.
I just had to get my Screen 1 a little more dialed in, and then work the strength slider. I still need to figure out all the available options for Luminance and Color modify and see how that compares with the full sliders in 1.8. The 1.9 gamma still seems to be global however, whereas when I tried the same in 1.8 it seems fully tied to the NB Accenting.
One other tidbit I noticed but haven't fully worked out yet - a late stage OptiDev re-do of global stretch in 1.9 seems to unravel deringing, which presumably was applied in SVD. I don't think I notice that with a late AutoDev re-do (post SVD) in 1.8.
Re: StarTools 1.9 preview
another thing I found,
artifacts on Entropy module, as seen in the screenshot.
with the same file it doesn't occur on 1.8 version. regards
artifacts on Entropy module, as seen in the screenshot.
with the same file it doesn't occur on 1.8 version. regards
Re: StarTools 1.9 preview
New alpha is up, change log as follows;
Spanish is more work in progress, as I've been trying to get one language to work reliably so that others are easy.
Nevertheless, for out German and Spanish friends, edit the new config file like so;
Clear skies and wishing you wonderful people all the best for 2023!
- Interpreting LIVETIME FITS keyword for exposure time detection
- Made tweaks to mask generation in decon module
- More translations
- Now using single config file, instead of separate files for each setting
- Fixed Auto Mask generator presets not turning off star mask generator
- Fixed some apodization mask generation failure cases
Spanish is more work in progress, as I've been trying to get one language to work reliably so that others are easy.
Nevertheless, for out German and Spanish friends, edit the new config file like so;
Lots more to do, but beta isn't too far off.ui_language=deutsch
Clear skies and wishing you wonderful people all the best for 2023!
Ivo Jager
StarTools creator and astronomy enthusiast
StarTools creator and astronomy enthusiast
-
- Posts: 1164
- Joined: Sun Jun 20, 2021 10:05 pm
- Location: Alta Loma, CA
Re: StarTools 1.9 preview
Sweet!
Um, gracias, and, danke schon.
Um, gracias, and, danke schon.
Re: StarTools 1.9 preview
I would've managed Dutch, but if it weren't for @hixx and @Carles "gracias" and "danke schön" would've indeed been my entire contribution to the Spanish and German translation too.
Ivo Jager
StarTools creator and astronomy enthusiast
StarTools creator and astronomy enthusiast
Re: StarTools 1.9 preview
Hi all,
As we are still in Alpha, I'd like to encourage german native speaking users to provide feedback on the german translation
Is it helpful to You?
Do You have any suggestions for changes?
Regards,
Jochen
As we are still in Alpha, I'd like to encourage german native speaking users to provide feedback on the german translation
Is it helpful to You?
Do You have any suggestions for changes?
Regards,
Jochen
Re: StarTools 1.9 preview
Dank u wel, Ivo (This is almost alI can write, but I can guess many more Dutch words when reading them )
Hi Jochen, thank you very much for your work on the translation. I'll install the new version, switch to German and have a look at it. I will PM you in case there's something striking me. (I guess the rest of the world is not interested in German Umlauts and Schiebereglers )
Best regards, Dietmar.
P.S.: Do you have some kind of text file containing all translated texts? If so and if you think it makes sense, you could send me this file via PM. (But having read parts of your German translation of the manual, I don't think this would be necessary at all.)