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

Pixinsight "Square Star" Issue

  • Please log in to reply
15 replies to this topic

#1 Juno18

Juno18

    Messenger

  • *****
  • topic starter
  • Posts: 459
  • Joined: 07 Dec 2018
  • Loc: Long Beach, Mississippi, USA

Posted 17 September 2021 - 05:03 PM

Good Day CN,

 

I just thought that I would share this information in case a new user of Pixinsight encounters the same issue that I did.

 

When I first started using PI about 9 months ago, after running the WBPP script to pre-process my data, the brighter stars were "square" or "pointed" looking.

It definitely looked like an anomaly to me. Possibly something wrong in my optical train. I was hesitant to post to the PI forum because of my "newness" to the software, so I used DSS to stack. In the DSS stacks, the same "square" stars that produced by PI, were nice and round. I did some researching on Google and saw that this issue could be remediated (somewhat)
by changing some settings in PI for pre-processing or just used DSS for pre-processing.

 

So, for the last 9 months, I have used the modified settings in Pixinsight’s WBPP.

 

Just recently, I started using a new scope (Sharpstar 61 EDPHII) with the same camera (ASI 533 MC-P) and reset the PI WBPP script settings back to default. When I generated the first masterlight with the new scope, I zoomed in and saw the same "square" stars again!

 

So, I ran a test so that I had the images to post to the PI forum.

 

I prepared a PI masterlight with the WBPP defaults (with "square" stars), another PI masterlight with the modified settings that I had been using (only slightly "square" stars) and an Autosave.fit with DSS to compare the three.

I applied a screen stretch to all three with no additional processing.

 

The WBPP masterlight with default settings had the "square" stars, but had obviously better quality (resolution) than the WBPP settings that I had been using trying to minimize the "square" star effect. The DSS autosave.fit had the nice round stars, but the resolution was the lowest of the three in the test.

 

I posted to the PI forum and here is the link in case you are interested.

 

https://pixinsight.c...re-stars.17172/

 

Here is the comparison image that I submitted to the PI forum (I don't think that it can be viewed unless you are logged into the forum). This is a 2x zoom in on a part of the M31 dust lanes with only an unlinked screen stretch with no processing.

 

PI Image Registration1.JPG

 

The PI WBPP with defaults is the image on the left (Square stars with the best resolution). The PI WBPP with the settings that I have been using to get "rounder" stars is in the middle. The DSS stack is on the right.

 

I found very little searching for this on Google or in the PI forum, but just in case you might have noticed a similar issue and were not sure why, here you go.

 

Ignorance on my part and I wish that I had posted the issue to the PI forum earlier, but definitely an unexpected happy ending!


  • sharkmelley likes this

#2 rgsalinger

rgsalinger

    Cosmos

  • *****
  • Moderators
  • Posts: 9,846
  • Joined: 19 Feb 2007
  • Loc: Carlsbad Ca

Posted 17 September 2021 - 05:43 PM

I think that your just way under sampled and so the stars are never going to be round if you've got good focus. It's really (IMHO) that simple. I'd be curious to know what settings produce rounder stars in WBPP, though, I didn't know that was possible to do. 

 

Rgrds-Ross


  • Juno18, SnowWolf and licho52 like this

#3 licho52

licho52

    Mariner 2

  • -----
  • Posts: 225
  • Joined: 15 Dec 2020

Posted 17 September 2021 - 06:03 PM

Drizzling makes my stars round while the normal integration process gives me pointy, blocky stars.  It's just undersampling.

 

Drizzle and downsample your pictures if you want the best quality.


  • sharkmelley, ks__observer and Juno18 like this

#4 sharkmelley

sharkmelley

    Fly Me to the Moon

  • *****
  • Posts: 5,004
  • Joined: 19 Feb 2013
  • Loc: UK

Posted 17 September 2021 - 06:05 PM

Thanks for that comparison.  Despite the "square" stars, the left hand image has sharper details than the other two.  The middle image shows that reducing the clamping threshold reduces some of that detail.  In the left image what I notice more than the square stars are the horizontal and vertical spikes on the saturated stars.

 

I think you should try drizzle integration, which will give the best of both worlds - sharp detail without the artifacts.

 

Mark


Edited by sharkmelley, 17 September 2021 - 06:06 PM.

  • idclimber and Juno18 like this

#5 Juno18

Juno18

    Messenger

  • *****
  • topic starter
  • Posts: 459
  • Joined: 07 Dec 2018
  • Loc: Long Beach, Mississippi, USA

Posted 17 September 2021 - 06:23 PM

I think that your just way under sampled and so the stars are never going to be round if you've got good focus. It's really (IMHO) that simple. I'd be curious to know what settings produce rounder stars in WBPP, though, I didn't know that was possible to do. 

 

Rgrds-Ross

Yes, I am under sampled with the Sharpstar scope. However, I saw the same effect at a Pixel scale: 1.356 arcsec/pixel with the ES scope. That is why I reduced the Clamping threshold to 0.1 and used Bicubic Spline pixel interpolation.

 

I see the same "square" star effects with drizzle integration. I have not tried downsampling though.

 

Thanks for that comparison.  Despite the "square" stars, the left hand image has sharper details than the other two.  The middle image shows that reducing the clamping threshold reduces some of that detail.  In the left image what I notice more than the square stars are the horizontal and vertical spikes on the saturated stars.

 

I think you should try drizzle integration, which will give the best of both worlds - sharp detail without the artifacts.

 

Mark

Absolutely! The left side image (even in the low quality snip) is by far better than the lower clamping threshold version and way better than the DSS stack. I see the same "square" star effect with drizzle integration, however, maybe I am not appling it correctly. I am simply checking the "generate drizzle data" box in the Lights tab, Subframe Weighing section.

 

I agree with your comment about the horizontal and vertical spikes on the saturated stars. Just reporting on the response by Juan on the PI forum.



#6 zakry3323

zakry3323

    Vanguard

  • *****
  • Posts: 2,198
  • Joined: 11 Apr 2016
  • Loc: Fairfield, Westmoreland County, PA

Posted 17 September 2021 - 06:29 PM

I'm simply checking the "generate drizzle data" box in the Lights tab, Subframe Weighing section.

 

You need to run the "Drizzle Integration" process using those drizzle files you've created.


  • rgsalinger, sharkmelley and Juno18 like this

#7 Juno18

Juno18

    Messenger

  • *****
  • topic starter
  • Posts: 459
  • Joined: 07 Dec 2018
  • Loc: Long Beach, Mississippi, USA

Posted 17 September 2021 - 07:18 PM

You need to run the "Drizzle Integration" process using those drizzle files you've created.

Thanks zakry.I will re-run WBPP with drizzle and try out that process. Much appreciated!


  • zakry3323 likes this

#8 zakry3323

zakry3323

    Vanguard

  • *****
  • Posts: 2,198
  • Joined: 11 Apr 2016
  • Loc: Fairfield, Westmoreland County, PA

Posted 17 September 2021 - 07:44 PM

Thanks zakry.I will re-run WBPP with drizzle and try out that process. Much appreciated!

Post results! 

 

Cheers!


  • Juno18 likes this

#9 jonnybravo0311

jonnybravo0311

    Mercury-Atlas

  • *****
  • Posts: 2,956
  • Joined: 05 Nov 2020
  • Loc: NJ, US

Posted 17 September 2021 - 07:59 PM

I think it appropriate to post this here... I did this a while back to show the effect of doing a 2x drizzle integration on some of my data. With my 294MM Pro running in its "default" mode, I'm at a 2.5"/px scale. The image below shows the standard integration on the left, 2x drizzle on the right:

 

gallery_347158_15202_14374.png


  • Juno18 and OldManSky like this

#10 Juno18

Juno18

    Messenger

  • *****
  • topic starter
  • Posts: 459
  • Joined: 07 Dec 2018
  • Loc: Long Beach, Mississippi, USA

Posted 17 September 2021 - 08:21 PM

I think it appropriate to post this here... I did this a while back to show the effect of doing a 2x drizzle integration on some of my data. With my 294MM Pro running in its "default" mode, I'm at a 2.5"/px scale. The image below shows the standard integration on the left, 2x drizzle on the right:

 

gallery_347158_15202_14374.png

Amazing comparison Johnny, Thanks!

 

 

Post results! 

 

Cheers!

Hi Zakry,

 

Awesome difference in the drizzled image. The drizzled image is on the left with the non-drizzled masterlight on the right. The stars look great! However, some of the stars have red or green dots in the center. Any idea what caused this?

 

drizzle comparison.JPG


  • zakry3323 likes this

#11 zakry3323

zakry3323

    Vanguard

  • *****
  • Posts: 2,198
  • Joined: 11 Apr 2016
  • Loc: Fairfield, Westmoreland County, PA

Posted 17 September 2021 - 10:22 PM

Sometimes when I expose to get a little more signal for my nebulosity my star cores get blown out. There is a repair script that you can use, I'll try to find it for you tomorrow. In the meantime this might be helpful: https://pixinsight.c...in-stars.14341/


  • Juno18 likes this

#12 zakry3323

zakry3323

    Vanguard

  • *****
  • Posts: 2,198
  • Joined: 11 Apr 2016
  • Loc: Fairfield, Westmoreland County, PA

Posted 17 September 2021 - 10:42 PM

Here it is, "Repair HSV Script" https://pixinsight.c...-stretch.12357/


  • Juno18 likes this

#13 sharkmelley

sharkmelley

    Fly Me to the Moon

  • *****
  • Posts: 5,004
  • Joined: 19 Feb 2013
  • Loc: UK

Posted 18 September 2021 - 01:02 AM

Awesome difference in the drizzled image. The drizzled image is on the left with the non-drizzled masterlight on the right. The stars look great! However, some of the stars have red or green dots in the center. Any idea what caused this?

Very odd.  I assume the centres of those stars are saturated in all 3 colour channels so it's difficult to guess what could be causing it.  The OP in the thread that zakry3323 linked suggests the Debayer process is causing this but I've never seen such an effect in my own processing of DSLR/Mirrorless images.  Even if Debayer causes this effect we would not expect to see the problem in a Drizzled image because Drizzle does not use the Debayered files - that's the whole point of Drizzle.

 

Do you have a raw file you can share?

 

Mark


Edited by sharkmelley, 18 September 2021 - 01:23 AM.

  • Juno18 likes this

#14 Juno18

Juno18

    Messenger

  • *****
  • topic starter
  • Posts: 459
  • Joined: 07 Dec 2018
  • Loc: Long Beach, Mississippi, USA

Posted 18 September 2021 - 06:31 AM

Here it is, "Repair HSV Script" https://pixinsight.c...-stretch.12357/

Thanks a bunch Zakry! 

 

I will give that a try soon. Got a bunch of running around today, so I probably won't to it until this evening. 

 

Just to note, I don't believe that I chose a Pixel Rejection in ImageIntegration. I will try that too.

 

 

Very odd.  I assume the centres of those stars are saturated in all 3 colour channels so it's difficult to guess what could be causing it.  The OP in the thread that zakry3323 linked suggests the Debayer process is causing this but I've never seen such an effect in my own processing of DSLR/Mirrorless images.  Even if Debayer causes this effect we would not expect to see the problem in a Drizzled image because Drizzle does not use the Debayered files - that's the whole point of Drizzle.

 

Do you have a raw file you can share?

 

Mark

I do Mark. Here is the link https://drive.google...iew?usp=sharing

 

By the way, you probably have noticed that this is a different image than the M31 images that I posted at the start of this thread and to the PI forum. This is Pickering's Triangle (or a piece of it), the first image that I shot with this scope/camera combination last month. I just happened to be reworking it. Same issue with the square stars as the M31 images.



#15 ChrisWhite

ChrisWhite

    Fly Me to the Moon

  • *****
  • Posts: 7,455
  • Joined: 28 Feb 2015
  • Loc: Northeast USA

Posted 18 September 2021 - 06:52 AM

Make sure clip high pixels and clip high range is not checked in imageintegration.


  • calypsob and Juno18 like this

#16 Juno18

Juno18

    Messenger

  • *****
  • topic starter
  • Posts: 459
  • Joined: 07 Dec 2018
  • Loc: Long Beach, Mississippi, USA

Posted 18 September 2021 - 01:50 PM

Thanks a lot for all of the suggestions!

 

I think that I found a workable configuration.

 

Here again, the image below is my original non-drizzled image (with square stars) on the left with the first drizzle attempt on the right. I did not chose a Pixel Rejection Algorithm in Image Integration. The drizzling was effective, however, some stars had a colored core.

 

Original STD image with no Drizzling.JPG

 

I then went back into Image Integration and chose the Generalized Extreme Studentized Deviate (ESD) Test Pixel Rejection Algorithm and unchecked clip high pixels. Clip high range was not checked in Image Integration (default). This is the image on the left below.

 

The colored star cores were gone, but there were clusters of red and green artifacts scattered throughout the image (image on the left below).

 

The image on the right (below) was with only the Generalized Extreme Studentized Deviate (ESD) Test Pixel Rejection Algorithm selected and clip high range was checked (default).

 

GESD with.JPG

 

The drizzled image on the right looks good to me. No colored star cores and no clusters of artifacts. The only change was using a Pixel Rejection Algorithm in Image Integration.

 

One question: There is a big difference between the stretched non-drizzle vs drizzled images. Non-drizzle on the left and drizzle on the right. Anything in particular that I need to watch out for in post processing?

 

non drizzle vs drizzle

 

 

Thank you so much for the help everyone!


  • zakry3323 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