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

Star Reduction/removal the classy way

  • Please log in to reply
6 replies to this topic

#1 Becomart

Becomart

    Apollo

  • -----
  • topic starter
  • Posts: 1,247
  • Joined: 03 Jan 2015

Posted 24 January 2021 - 06:31 PM

A good night of imaging for me usually yields some strong data and well exposed star fields but  my images are often littered with tiny stars across the image. Whilst this is clearly dependent on the target, I find the best imagers don’t allow small artefact stars to bleed through in a final image. Their images are smooth with controlled prominent stars and little to no domination. 
Using the usual morphological transformation tools and other techniques leads to less stars at the cost of artefacts or, worse still, a pitted background.

 

So, what’s the secret? How do the best achieve the controlled look. How do you go from a look like this https://www.astrobin...02650/B/?nc=all to a look like this https://www.astrobin.com/sopidb/0/?nc=all

I was working on this object today and my current effort is probably a halfway point between the two extremes. 


Edited by Becomart, 24 January 2021 - 06:33 PM.


#2 B 26354

B 26354

    Surveyor 1

  • *****
  • Posts: 1,840
  • Joined: 05 Jan 2017
  • Loc: Southern California semi-desert (NELM mag 5.3)

Posted 24 January 2021 - 06:39 PM

2nd link doesn't work.  grin.gif



#3 limeyx

limeyx

    Viking 1

  • -----
  • Posts: 857
  • Joined: 23 May 2020
  • Loc: Seattle, WA

Posted 24 January 2021 - 06:52 PM

Do you want reduction or removal ?

 

This works well for me to reduce stars

 

https://www.youtube....h?v=3DdSDoJfjM4



#4 Cbaxter

Cbaxter

    Explorer 1

  • -----
  • Posts: 91
  • Joined: 29 Mar 2020

Posted 24 January 2021 - 11:18 PM

What post-processing software do you use? Sorry if it's in your signature. Can't see signatures on my mobile.

Quick note on the referenced images; the first link is an LRGB image while the second is a narrowband image. The second image would have much smaller/dimmer stars to begin with, then the editor also did star reduction on it. Honestly, to me personally, that second image looks way off because of all the artifacts where small stars used to be. *EDIT* After more closely looking at the full size version of that second image you linked, it does not look like star removal artifacts but rather just the stars have been dimmed a lot so they kind of blend with the background. It doesn't look good to me though.

I used to mostly use star masks and StarTools/GIMP (PI probably has multiple ways to mask and manipulate stars) lately I have been using StarNet++ a lot. I set up NVIDIA CUDA GPU acceleration with StarNet++ which makes the process significantly faster.

With StarNet I typically take the starless image and subtract it from a copy of the original, giving me a layer of just stars. Then I can process the stars and nebula separately, giving a lot of control over how the stars look.

StarTools can generate star masks then manipulate just the stars (invert mask for the reverse) with the available tools, including star shrink/tighten. The latest version of ST also supports GPU acceleration which makes star mask generation nearly instantaneous. Super Structure tool in ST is also useful in dimming small stars or isolating structures like nebula.

Regards.

Edited by Cbaxter, 24 January 2021 - 11:30 PM.


#5 Becomart

Becomart

    Apollo

  • -----
  • topic starter
  • Posts: 1,247
  • Joined: 03 Jan 2015

Posted 25 January 2021 - 03:30 PM

What post-processing software do you use? Sorry if it's in your signature. Can't see signatures on my mobile.

Quick note on the referenced images; the first link is an LRGB image while the second is a narrowband image. The second image would have much smaller/dimmer stars to begin with, then the editor also did star reduction on it. Honestly, to me personally, that second image looks way off because of all the artifacts where small stars used to be. *EDIT* After more closely looking at the full size version of that second image you linked, it does not look like star removal artifacts but rather just the stars have been dimmed a lot so they kind of blend with the background. It doesn't look good to me though.

I used to mostly use star masks and StarTools/GIMP (PI probably has multiple ways to mask and manipulate stars) lately I have been using StarNet++ a lot. I set up NVIDIA CUDA GPU acceleration with StarNet++ which makes the process significantly faster.

With StarNet I typically take the starless image and subtract it from a copy of the original, giving me a layer of just stars. Then I can process the stars and nebula separately, giving a lot of control over how the stars look.

StarTools can generate star masks then manipulate just the stars (invert mask for the reverse) with the available tools, including star shrink/tighten. The latest version of ST also supports GPU acceleration which makes star mask generation nearly instantaneous. Super Structure tool in ST is also useful in dimming small stars or isolating structures like nebula.

Regards.

You got me on the examples...it was late last night and I was trying to find quick examples to illustrate what I was talking about. I am using Pixinsight and also use starnet (with the cuda acceleration) but find this leaves artifacts. I guess I'll return to this thread at some point with my own examples which might make more sense.



#6 Rasfahan

Rasfahan

    Messenger

  • -----
  • Posts: 419
  • Joined: 12 May 2020
  • Loc: Hessen, Germany

Posted 25 January 2021 - 04:33 PM

One thing is that your first link is with a 16" RC and the second one with an 8" newtonian. Stars gain in brightness proportional to the square of the aperture (this is different from diffuse light sources), so a 16" telescope will have much brighter stars than an 8" one. In addition, the image scale is different: The first image has 0.5 arcsec/pixel, the second one is 1.0 arcsec/pixel, so "same size" stars would seem half the size in the second one. Then RC scopes tend to have a bit larger stars than newtonians, due to a larger diffraction pattern by the higher central obstruction. And then there is, most importantly as already stated, narrowband vs LRGB: So, there are four good reasons already why the first image has more prominent stars.  

 

Also, I, personally, do not like the stars in the second image better than the first: They have obvious processing artefacts: Ringing, the shapes of larger stars are off, and most importantly, the diffraction spikes of the newtonian are mostly missing. I think the second picture looks a lot like starnet++ was used to fabricate a grown/blurred mask, and differential stretch between stars and nebula - actually I now realise that is basically what the author did in Photoshop (he stretched first and then subtracted). 

 

In Pixinsight, I found the EZ Processing Suite has a quite decent star reduction algorithm. It basically semi-automates what I did by hand before, and I like the results in some images (especially my 11" RASA images seem to profit). 

 

In general, I feel star reduction techniques should be used very sparingly, because they tend to cause the aforementioned artefacts. 

 

The first picture is five years old, long before starnet++ existed. I am not sure any star reduction techniques were used. I also do not think the image needs it. Personally, I like it very much - but that is a matter of taste. 


  • Cbaxter likes this

#7 Becomart

Becomart

    Apollo

  • -----
  • topic starter
  • Posts: 1,247
  • Joined: 03 Jan 2015

Posted 25 January 2021 - 05:40 PM

As said...I used poor examples to illustrate my point. Taste but I'm not keen on either image as an overall effect but I do like the actual crescent nebula in the second one and my version I am working on is not too dissimilar. I do like the star reduction in the scripts, especially the Adam Block technique. 


Edited by Becomart, 25 January 2021 - 05:40 PM.



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