Jump to content

  •  

CNers have asked about a donation box for Cloudy Nights over the years, so here you go. Donation is not required by any means, so please enjoy your stay.

Photo

AutoStakkert - "Quality" value meaning?

  • Please log in to reply
88 replies to this topic

#1 RedLionNJ

RedLionNJ

    Fly Me to the Moon

  • *****
  • Moderators
  • topic starter
  • Posts: 6,556
  • Joined: 29 Dec 2009
  • Loc: Red Lion, NJ, USA

Posted 04 October 2022 - 10:47 AM

Hi,

 

I've been using AutoStakkert since around the time Emil originally released it. And during that time, I've been ignoring the numbers in the top left of the display panel:

 

aspanel.png

 

I saw someone mention (sorry, I forget who, now) the other day he was seeing quality values in six digits (like, over 100,000).  I rarely see anything more than 3 digits (like the 137 in this example).  And the data that produced this 137 was perfectly good, definitely some of my better OSC capture data.

 

Does anybody know the meaning of this "Q:" value and the math underlying it?

 

Thanks,

 

Grant

 

P.S. to other mods - since this is so specific to planetary imaging, I'd like it left here, please.

 


  • happylimpet, gfstallin and moonwatching ferret like this

#2 Borodog

Borodog

    Fly Me to the Moon

  • *****
  • Posts: 6,983
  • Joined: 26 Oct 2020

Posted 04 October 2022 - 01:03 PM

I assume it's the integral of a Laplacian quality estimator over the entire frame. It will vary from image to image depending on the target and the image scale, the Noise Robust, the darkness of the sky background, etc., ignoring the variability due to quality itself. You can't compare your own quality numbers for Jupiter to Saturn, much less to somebody else's for a completely different rig and settings.


  • RedLionNJ, happylimpet and KiwiRay like this

#3 KiwiRay

KiwiRay

    Mercury-Atlas

  • -----
  • Posts: 2,552
  • Joined: 09 Apr 2015
  • Loc: Seattle, USA

Posted 04 October 2022 - 01:28 PM

I'd always ignored those values until a few days ago when I was trying to find a place to cut off a series of poor frames. My good values were all in the 20s, and while I'd always considered AS!3 quality to be relative when using the graph, I did wonder if this was supposed to be some sort of absolute measure. Clearly not, and thanks Borodog for helping explain why.


  • RedLionNJ likes this

#4 yock1960

yock1960

    Fly Me to the Moon

  • *****
  • Posts: 5,439
  • Joined: 22 Jun 2008
  • Loc: (Crossroad of clouds) Ohio, USA

Posted 04 October 2022 - 03:00 PM

How they are determined? Who knows...apart from Emil. I actually like the values in AS!2 better, since it gives a number for quality and a brightness. Brightness seems to affect quality score, so while a brighter image may 'score' well, it may actually yield a poorer image. I do my best to maintain a constant brightness and use the quality values to tweak focus to perfection....usually failing....but it does help, especially if you keep your collimation reasonably good throughout a session....which I don't always do...shrug.gif....falling quality scores with increasing elevation is often a sign that your collimation has fallen off.  

 

Steve


Edited by yock1960, 04 October 2022 - 03:01 PM.


#5 RedLionNJ

RedLionNJ

    Fly Me to the Moon

  • *****
  • Moderators
  • topic starter
  • Posts: 6,556
  • Joined: 29 Dec 2009
  • Loc: Red Lion, NJ, USA

Posted 04 October 2022 - 03:13 PM

I do my best to maintain a constant brightness and use the quality values to tweak focus to perfection....usually failing....but it does help, especially if you keep your collimation reasonably good throughout a session....which I don't always do...shrug.gif....falling quality scores with increasing elevation is often a sign that your collimation has fallen off.  

 

Steve

Thanks, Steve. But this is all in post-capture processing, so are you "test stacking" at the scope to try to maximize the quality measure in AutoStakkert?



#6 Borodog

Borodog

    Fly Me to the Moon

  • *****
  • Posts: 6,983
  • Joined: 26 Oct 2020

Posted 04 October 2022 - 03:14 PM

I'd always ignored those values until a few days ago when I was trying to find a place to cut off a series of poor frames. My good values were all in the 20s, and while I'd always considered AS!3 quality to be relative when using the graph, I did wonder if this was supposed to be some sort of absolute measure. Clearly not, and thanks Borodog for helping explain why.

 

The absolute quality values can be critical in my opinion. Just know that they are not comparable across different targets, cameras, etc.

 

 

How they are determined? Who knows...apart from Emil. I actually like the values in AS!2 better, since it gives a number for quality and a brightness. Brightness seems to affect quality score, so while a brighter image may 'score' well, it may actually yield a poorer image. I do my best to maintain a constant brightness and use the quality values to tweak focus to perfection....usually failing....but it does help, especially if you keep your collimation reasonably good throughout a session....which I don't always do...shrug.gif....falling quality scores with increasing elevation is often a sign that your collimation has fallen off.  

 

Steve

 

It's pretty straightforward. It will be something like this: https://pyimagesearc...on-with-opencv/

 

You convolve the image with the Laplacian kernel, and then add up all the resulting values for the image. The Noise Robust setting bins the image before this is done. An NR of 6 will bin the image 6x6, for example. Higher values are interpreted as higher quality. That is indeed why brighter frames are marked as higher quality. That is one good time to use PIPP for pre-processing, when you have variable transparency. You can have PIPP stretch the frames so that the dynamic range matches.
 


  • yock1960, RedLionNJ, happylimpet and 3 others like this

#7 yock1960

yock1960

    Fly Me to the Moon

  • *****
  • Posts: 5,439
  • Joined: 22 Jun 2008
  • Loc: (Crossroad of clouds) Ohio, USA

Posted 04 October 2022 - 03:26 PM

Thanks, Steve. But this is all in post-capture processing, so are you "test stacking" at the scope to try to maximize the quality measure in AutoStakkert?

Absolutely!



#8 Ittaku

Ittaku

    Mercury-Atlas

  • *****
  • Posts: 2,866
  • Joined: 09 Aug 2020
  • Loc: Melbourne, Australia

Posted 04 October 2022 - 03:43 PM

The absolute quality values can be critical in my opinion. Just know that they are not comparable across different targets, cameras, etc.

This is by far the most important point. The object you are capturing and even the gain also affects it, so you can't compare captures purely on the absolute value unless everything else is kept equal. A common problem is using autohistogram during varying high cloud cover - as the transparency drops and FC raises the gain to compensate, the image gets grainier. Grain is interpreted as contrast in Autostakkert, so if you had say a 20 minute long capture and tried to find the best value and section of the capture, you will unwittingly end up choosing the lowest transparency times.


  • happylimpet likes this

#9 Tulloch

Tulloch

    Fly Me to the Moon

  • *****
  • Posts: 5,002
  • Joined: 02 Mar 2019
  • Loc: Melbourne, Australia

Posted 04 October 2022 - 04:30 PM

Last year I did a comparison between different framerates on Jupiter to see if this made a difference to my captures. I captured from 30fps up to 200fps and found very little difference between the final results. However, my gain settings were obviously quite different between captures, but the "best" quality values I measured were quite similar. Here are the results in very good seeing:

 

fps     Quality     Gain

30      85425       150

100    87008       250

150    90100       280

200    84281       305

 

I also compared some more recent Jupiter videos I had, one from a week ago in good seeing and another a month ago in poor seeing. Same equipment, framerate and gain as above, but the quality values are significantly different.

 

fps     Quality     Seeing

150    52300      Good

150    36893      Poor

 

My results show that the maximum quality number may not necessarily be related to the amount of noise in the image, and are related to the "quality" of the video, whatever that is, possibly edge constrast?  It is interesting to see the large discrepency between what I called "average" seeing last year and "good" seeing this year - clearly, it's been a poor season down here smile.gif.

 

Some of the screen grabs attached below.

 

30fps

AS3 Jupiter quality 30fps.JPG

 

200 fps

AS3 Jupiter quality 200fps.JPG

 

"Good" seeing

AS3 Jupiter quality good seeing.JPG

 

"Poor" seeing

AS3 Jupiter quality poor seeing.JPG

 

 

Andrew


Edited by Tulloch, 04 October 2022 - 04:33 PM.

  • Applal and RedLionNJ like this

#10 RedLionNJ

RedLionNJ

    Fly Me to the Moon

  • *****
  • Moderators
  • topic starter
  • Posts: 6,556
  • Joined: 29 Dec 2009
  • Loc: Red Lion, NJ, USA

Posted 04 October 2022 - 04:42 PM

I know we shouldn't be comparing different scopes/exposures/gains, etc. - but it does still concern me many people are getting quality numbers in the tens (or even hundreds) of thousands, whereas mine tend to lurk in the low hundreds. Yet the final images are apparently ok.

 

For example:  quality 129.0

 

2022-09-10-0637_0-GB_r2.jpg

 

 

Go figure.


  • gfstallin, RickD_99 and firmament1 like this

#11 dcaponeii

dcaponeii

    Fly Me to the Moon

  • *****
  • Posts: 5,869
  • Joined: 01 Sep 2019
  • Loc: Waxahachie, TX

Posted 04 October 2022 - 04:54 PM

Are you using AS!3?  In the onecthat has been seeing values lately above 100,000 and my previous work has always been in the 30,000 - 50,000 range. In my case these new higher numbers have definitely aligned with better images.  I've never seen numbers in the hundreds across now 4 different cameras.  There must be some additional factor going on.



#12 Ittaku

Ittaku

    Mercury-Atlas

  • *****
  • Posts: 2,866
  • Joined: 09 Aug 2020
  • Loc: Melbourne, Australia

Posted 04 October 2022 - 05:24 PM

I know we shouldn't be comparing different scopes/exposures/gains, etc. - but it does still concern me many people are getting quality numbers in the tens (or even hundreds) of thousands, whereas mine tend to lurk in the low hundreds. Yet the final images are apparently ok.

 

For example:  quality 129.0

 

attachicon.gif2022-09-10-0637_0-GB_r2.jpg

 

 

Go figure.

Do you take wide field images with a larger sensor? My exact same object with a larger ROI in the exact same conditions shows dramatically lower quality values than the small ROI captures.


  • Tulloch likes this

#13 Tulloch

Tulloch

    Fly Me to the Moon

  • *****
  • Posts: 5,002
  • Joined: 02 Mar 2019
  • Loc: Melbourne, Australia

Posted 04 October 2022 - 05:59 PM

Do you take wide field images with a larger sensor? My exact same object with a larger ROI in the exact same conditions shows dramatically lower quality values than the small ROI captures.

Excellent point - here's two screengrabs from the same capture, one widefield and the other cropped in AS!3 to just include the planet.

 

fps     Quality     FOV

150    23257      Wide 

150    64236      Narrow

 

Same capture, different FOV in AS!3.

 

They are still much much higher than the q-factor in the hundreds with other cameras/setups, but FOV certainly makes a difference.

 

Andrew

 

Wide

AS3 Jupiter quality wide field.JPG

 

Narrow

AS3 Jupiter quality narrow field.JPG


Edited by Tulloch, 04 October 2022 - 05:59 PM.

  • Ittaku likes this

#14 Ittaku

Ittaku

    Mercury-Atlas

  • *****
  • Posts: 2,866
  • Joined: 09 Aug 2020
  • Loc: Melbourne, Australia

Posted 04 October 2022 - 06:07 PM

Last year I did a comparison between different framerates on Jupiter to see if this made a difference to my captures. I captured from 30fps up to 200fps and found very little difference between the final results. However, my gain settings were obviously quite different between captures, but the "best" quality values I measured were quite similar. Here are the results in very good seeing:

 

fps     Quality     Gain

30      85425       150

100    87008       250

150    90100       280

200    84281       305

 

I also compared some more recent Jupiter videos I had, one from a week ago in good seeing and another a month ago in poor seeing. Same equipment, framerate and gain as above, but the quality values are significantly different.

 

fps     Quality     Seeing

150    52300      Good

150    36893      Poor

 

My results show that the maximum quality number may not necessarily be related to the amount of noise in the image, and are related to the "quality" of the video, whatever that is, possibly edge constrast?  It is interesting to see the large discrepency between what I called "average" seeing last year and "good" seeing this year - clearly, it's been a poor season down here smile.gif.

 

Some of the screen grabs attached below.

I think this is different to when gain varies with transparency, but it's good to confirm when the transparency remains the same the quality algorithm isn't affected by gain. Also funny you said 30fps wasn't really different ;)


Edited by Ittaku, 04 October 2022 - 06:07 PM.


#15 Ittaku

Ittaku

    Mercury-Atlas

  • *****
  • Posts: 2,866
  • Joined: 09 Aug 2020
  • Loc: Melbourne, Australia

Posted 04 October 2022 - 06:09 PM

I think this is different to when gain varies with transparency, but it's good to confirm when the transparency remains the same the quality algorithm isn't affected by gain.

I actually have an example right here. There was precisely one passing cloud, and I'd left autohisto on. See if you can spot it.

 

Screenshot 20221005 100812

  • happylimpet likes this

#16 RedLionNJ

RedLionNJ

    Fly Me to the Moon

  • *****
  • Moderators
  • topic starter
  • Posts: 6,556
  • Joined: 29 Dec 2009
  • Loc: Red Lion, NJ, USA

Posted 04 October 2022 - 07:02 PM

Do you take wide field images with a larger sensor? My exact same object with a larger ROI in the exact same conditions shows dramatically lower quality values than the small ROI captures.

Nope. As small an ROI as I can practically get - 656 wide by 424 tall. Jupiter's globe takes up about 360 of that height.



#17 RedLionNJ

RedLionNJ

    Fly Me to the Moon

  • *****
  • Moderators
  • topic starter
  • Posts: 6,556
  • Joined: 29 Dec 2009
  • Loc: Red Lion, NJ, USA

Posted 04 October 2022 - 07:10 PM

Here's my full-screen AS!3 shot, by comparison. This one made it all the way up to 183, but produced the image in post #10, a few posts above.

 

asexample.jpg


  • Tulloch likes this

#18 Borodog

Borodog

    Fly Me to the Moon

  • *****
  • Posts: 6,983
  • Joined: 26 Oct 2020

Posted 04 October 2022 - 07:58 PM

I have to say, that is very weird. 


  • RedLionNJ likes this

#19 Tulloch

Tulloch

    Fly Me to the Moon

  • *****
  • Posts: 5,002
  • Joined: 02 Mar 2019
  • Loc: Melbourne, Australia

Posted 04 October 2022 - 08:08 PM

Here's my full-screen AS!3 shot, by comparison. This one made it all the way up to 183, but produced the image in post #10, a few posts above.

 

I note you are using version 3.0.14, while I'm using 3.1.4.

 

Maybe that's the difference?


  • RedLionNJ, Ittaku and Borodog like this

#20 Ittaku

Ittaku

    Mercury-Atlas

  • *****
  • Posts: 2,866
  • Joined: 09 Aug 2020
  • Loc: Melbourne, Australia

Posted 04 October 2022 - 08:23 PM

I note you are using version 3.0.14, while I'm using 3.1.4.

 

Maybe that's the difference?

I was comparing windows trying to find the difference and just spotted that. I'd say it's likely.


  • RedLionNJ and Tulloch like this

#21 Borodog

Borodog

    Fly Me to the Moon

  • *****
  • Posts: 6,983
  • Joined: 26 Oct 2020

Posted 04 October 2022 - 08:34 PM

Good catch. There is definitely no reason not to be using the latest version.


  • RedLionNJ likes this

#22 RedLionNJ

RedLionNJ

    Fly Me to the Moon

  • *****
  • Moderators
  • topic starter
  • Posts: 6,556
  • Joined: 29 Dec 2009
  • Loc: Red Lion, NJ, USA

Posted 05 October 2022 - 08:39 AM

Agreed to the last three posts!  I (obviously!) didn't notice that. It looks like I have 3.1.4 on my observatory computer, but 3.0.14 on the one in my home office.  Updating as we speak! Results to follow. :)



#23 RedLionNJ

RedLionNJ

    Fly Me to the Moon

  • *****
  • Moderators
  • topic starter
  • Posts: 6,556
  • Joined: 29 Dec 2009
  • Loc: Red Lion, NJ, USA

Posted 05 October 2022 - 08:48 AM

Well, that made the quality score LOWER (same data):

 

as314.jpg

 

 

Really peculiar frown.gif

 

 


  • Borodog likes this

#24 Borodog

Borodog

    Fly Me to the Moon

  • *****
  • Posts: 6,983
  • Joined: 26 Oct 2020

Posted 05 October 2022 - 08:53 AM

Who ordered that?


  • RedLionNJ likes this

#25 dcaponeii

dcaponeii

    Fly Me to the Moon

  • *****
  • Posts: 5,869
  • Joined: 01 Sep 2019
  • Loc: Waxahachie, TX

Posted 05 October 2022 - 11:32 AM

That is just the weirdest thing.  In my worst images I've still had quality factors above 10,000.  Do you set your AP's before you do analysis?  If I recall if there are not any AP's set then the Analysis run is done "global".  Could that be what your showing us?  What does the quality factor display after you've finished stacking.  In my case that number is always different (usually higher) than the result from analysis.


  • RedLionNJ likes this


CNers have asked about a donation box for Cloudy Nights over the years, so here you go. Donation is not required by any means, so please enjoy your stay.


Recent Topics






Cloudy Nights LLC
Cloudy Nights Sponsor: Astronomics