StarXterminator and removing stars

Comments

• BlockHead

  February 2022

  In the linear state you can add images together using PixelMath and maintain the true signal strengths and relationships between the data. If you permanently stretch the data...all of that information is lost. 

  This isn't about NB vs broadband- it is about whether you want to retain relationships between images- this is faithfully maintained in linear data.

  
  

  -the Blockhead  
• Alan

  February 2022

  Hi Adam

  
  

  Thanks for the reply - I understand your point about use of Pixel Math and linear images however, my question is not related to that fact. To explain: let's say you followed the alternative process:

  
  

  a) Take your linear narrowband data, decide on how you want to map these to the RGB colour space and  create a linear RGB image. 

  
  

  b) stretch the RGB image to create a non-linear RGB image.

  
  

  c) create a starless image of the narrowband data from the RGB image.

  
  

  d) reduce glows from stars

  
  

  e) take your RGB starfield data and stretch it until your happy with how the stars look

  
  

  f) screen the non-linear RGB starfield with the non-linear RGB starless image. 

  
  

  I was really interested in your thoughts about in this non-linear approach to starless image processing compared to the linear equivalent.  

  
  

  Alan

  
  

  
  
• Anne van Houwelingen

  February 2022 edited February 2022

  Hi Alan,

  
  

  I am (no) BlockHead, but in the spirit of interaction I will give you my view.

  
  

  In narrowband imaging, it is quite customary that signal strengths are taken out of proportion. Specifically the OIII signal (when doing HOO) is much weaker than the Ha signal. What I do is to make starless nonlinear versions of both Ha and OIII, boosting OIII to my liking (and what the noise level wil allow). Then choose your favourite method of combining the channels and see what comes out and experiment with either the formulae or the OIII strength. Or both. It is just art and representation.

  
  

  So, I would do steps a and b differently.

  
  

  A few other remarks:

  1. StarXterminator (and both versions of Starnet) behave a little better or worse in lin vs non-lin. You just have to experiment with that. As I use non-linear images before combining in NB anyway, for me it is just a matter of finding what gives the cleanest result for that particular image.
  2. You might want to add instead of screen for adding back stars, especially if there are stars on top of bright nebulosity. In the process of de-starring, the ratio of the starintensities is altered. Have a look on the generated starmask and compare to the original. When you screen this mismatch is preserved. When you add the ratios are restored again.

  CS, Anne

  
  
• Alan

  February 2022

  Hi Anne

  
  

  Thanks for your comments. 

  
  

  Yes, your method is actually what I've currently been doing in my limited experience of StarXterminator. I started on this down this route based on my workflow experience from Photoshop where everything needed to be non-linear in order to make objects visible. 

  
  

  More recently, within Pixinsight,  I've been experimenting with keeping things linear whilst I see what the resultant RGB image looks like whilst I adjust the mapping of the colours. 

  
  

  On your point about adding stars back in -  In the above I was referring to a non-linear RGB starfield (acquired with broadband filters) which I suggested could be Screened with the resultant non-linear RGB narrowband image.  I'm not referring to the starfield generated from StarXterminator which is a by-product of the producing the starless image - I agree that these will needed to be added in. 

  
  

  Anyway, thanks for your comments. I think I need to experiment with StarXterminator a bit more !

  
  

  Alan
• Anne van Houwelingen

  February 2022

  Hi Alan,

  
  

  Do not forget to include Starnet V2 in your experiments. It has been much improved.

  
  

  Sometimes StarXterminator works better, sometimes Starnet V2.

  
  

  
  

  Anne