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

Approaches to dealing with Alnitak--how many ways? any one superior?

  • Please log in to reply
24 replies to this topic

#1 ntph

ntph

    Viking 1

  • -----
  • topic starter
  • Posts: 743
  • Joined: 20 Oct 2012
  • Loc: Sudbury, ON

Posted 06 February 2021 - 05:47 PM

This area probably deserves all the attention it gets from astrophotographers. Bright nebulae, dark nebulae, hydrogen-alpha, colourful dust and stars—it has it all, including Alnitak and all the challenges that having such a bright star in such an environment gives us. This is my first serious attempt at imaging this part of the sky. I have been reading about various ways to deal with reflections or haloes and decided I would try at least one. I chose Charlie Bracken’s method of creating a mask using PixelMath and a clone of the image, as he describes in his website tutorial:

 

https://digitalstars...-in-pixinsight/

 

With the mask applied, you adjust the brightness of the halo on your target image using HistogramTransformation, moving the lefthand slider up to taste. I did this process three times, making a new mask each time, and being careful not to clip “too much". It gave me a smoother result than trying to do things all at once, but I wasn’t as “scientific” as I could have been. I suspect that if the size of the next mask created is decreased by the same amount as the convolution applied previously, it may yield a better result. I will give it a try, but for now I am fairly happy with the result.

 

Prior to reduction:

HHF_preglareredn copy.jpg

 

After three iterations:

Flame and Horse copy.jpg

 

I am wondering where in the process is best to apply this. I did it once a “final” image was created (nonlinear, noise reduced, sharpened, colour saturation). I am also wondering if anyone has compared various methods of halo reduction?There are several to try and I will try to find the time to give other methods a try, but thought I would throw it out there to see if anyone cares to comment.

 

Sadly, whlle it doesn’t show in this image, my oil-spaced objective has developed a serious problem and is on its way back to Astro-Physics for Roland and co. to diagnose and repair. i purchased it in 1996 when the waiting list was only about 6 months; twenty-five years of service is pretty good, but a bit concerning that this happened at all. The only possible cause I can come up with is prolonged exposure to moderately (by Canadian, or high elevation, standards) cold temperatures (-15 to -25C). There is a separate post on this problem.

 

Details on images:  first is pre-halo reduction, second is post (hopefully obvious!)…. 49 x180 seconds QHY168c at -20C camera temp with no cooling power applied! AstroPhysics EDF 130 mm f/6 on CGEM, PHD2, NINA, PI processing (WBPP, EZDecon, EZNoiseReduction, HT/CT/MLTsharpening, LHE then halo experiments)


  • mikewayne3 likes this

#2 imtl

imtl

    Fly Me to the Moon

  • *****
  • Moderators
  • Posts: 5,895
  • Joined: 07 Jun 2016
  • Loc: Down in a hole

Posted 06 February 2021 - 06:08 PM

You can do this with StarHaloReducer script in PI. Instead of masking iterations by hand as you are doing now.



#3 ntph

ntph

    Viking 1

  • -----
  • topic starter
  • Posts: 743
  • Joined: 20 Oct 2012
  • Loc: Sudbury, ON

Posted 06 February 2021 - 06:41 PM

Eyal, I have tried to figure it out, but the documentation is scarce and I couldn't get it to work properly. User error, obviously, and more persistence needed but better documentation would be helpful....



#4 imtl

imtl

    Fly Me to the Moon

  • *****
  • Moderators
  • Posts: 5,895
  • Joined: 07 Jun 2016
  • Loc: Down in a hole

Posted 06 February 2021 - 07:43 PM

Norm, I can show you a few tricks on how to get it to work. Let's pm when you want to.


  • ntph likes this

#5 ntph

ntph

    Viking 1

  • -----
  • topic starter
  • Posts: 743
  • Joined: 20 Oct 2012
  • Loc: Sudbury, ON

Posted 13 February 2021 - 07:59 PM

Eyal and I agreed that maybe this wasn't the best halo to tackle with the StarHaloReducer script!

 

But I did add about 3 hours worth of Ha data extracted from a series of L-eNhance filter subs and had a go at an Ha-RBG image, a first for me. It certainly adds intensity to the red part of the spectrum! I did process the Ha data before combining it (deconvolve, denoise, stretch, sharpen) and then fine tuned the resultant combination image, but it could still use more of a lot of stuff, like integration time!  

 

Getting good background extraction from a field that basically has no "normal" background is something I don't think I have really figured out how to do. Now there's a topic I might post separately, if no one jumps in on this (which I don't really expect). 

 

HaRGB_Flame_Horsehead copy.jpg

 

 


  • F.Meiresonne, mikewayne3 and imtl like this

#6 imtl

imtl

    Fly Me to the Moon

  • *****
  • Moderators
  • Posts: 5,895
  • Joined: 07 Jun 2016
  • Loc: Down in a hole

Posted 13 February 2021 - 08:09 PM

Norm, I think the image is darn good. Can you send me a link to the XISF final file? I want to try something out on this halo.



#7 Stelios

Stelios

    Voyager 1

  • *****
  • Posts: 11,127
  • Joined: 04 Oct 2003
  • Loc: West Hills, CA

Posted 13 February 2021 - 09:17 PM

The GAME script beats the pants out of Star Halo Reducer (IMO). 


  • imtl likes this

#8 BQ Octantis

BQ Octantis

    Fly Me to the Moon

  • *****
  • Posts: 5,806
  • Joined: 29 Apr 2017
  • Loc: Mid Atlantic, USA

Posted 14 February 2021 - 04:31 AM

One of the clues to how to deal with anything we want to suppress is to examine the channels individually. Splitting out your channels, Alnitak and its bloom increases with decreasing wavelength—it has the least in the red channel, a lot in the green channel, and it blows out the blue channel:

 

gallery_273658_12412_361756.jpg

 

So we can just use the red channel as a gradated luminosity channel: the brighter the luminosity, the more we use the luminosity from the red channel. With Photoshop, I switch the image mode to Luminosity-a-b (Lab), duplicate the layer, and paste in the red channel for the L channel. I then add a layer mask and paste in the L channel from the original in as the mask. If I do the same with a color layer where I desaturate the blues and cyans, I can get a workable result:

 

gallery_273658_12412_93101.jpg

Animated toggle

 

The highlights in all the blues are quite clipped, so you might inspect your lights. If the linear data isn't blown out, you might try a different stretch on the blue channel that preserves the detail in the highlights.

 

Sorry I don't speak PI…

 

BQ


Edited by BQ Octantis, 14 February 2021 - 04:56 AM.

  • ntph, CharLakeAstro, imtl and 2 others like this

#9 BQ Octantis

BQ Octantis

    Fly Me to the Moon

  • *****
  • Posts: 5,806
  • Joined: 29 Apr 2017
  • Loc: Mid Atlantic, USA

Posted 14 February 2021 - 06:33 AM

Here's another crack at it with an entirely different approach. I built the luminosity out of screen layers of just the red and green channels. I then twiddled the curves on the green screen layer to reduce the halo. I used the original as a color layer and then just reduced the saturation of the blue and cyan.

 

try2.jpg

 

This better preserved the color and luminosity of the original in the horse's head. I preferred the color around Alnitak with this method, but I preferred the color in NGC 2023 with the prior method.

 

BQ


Edited by BQ Octantis, 14 February 2021 - 06:35 AM.

  • ntph likes this

#10 Tapio

Tapio

    Fly Me to the Moon

  • -----
  • Posts: 6,061
  • Joined: 24 Sep 2006
  • Loc: Tampere, Finland

Posted 14 February 2021 - 06:48 AM

I like BQ's versions.

#11 TxStars

TxStars

    Mercury-Atlas

  • *****
  • Posts: 2,921
  • Joined: 01 Oct 2005
  • Loc: Lost In Space

Posted 14 February 2021 - 06:57 AM

You know there is a problem child so perhaps control it a bit from the start?

Try using data obtained with shorter subs? ( 20 / 30 / 60 seconds)


Edited by TxStars, 14 February 2021 - 06:57 AM.

  • ntph likes this

#12 jonnybravo0311

jonnybravo0311

    Soyuz

  • *****
  • Moderators
  • Posts: 3,819
  • Joined: 05 Nov 2020
  • Loc: NJ, US

Posted 14 February 2021 - 11:35 AM

Sorry I don't speak PI…

 

BQ

Blasphemy!

 

I wrote this in another thread for a potential way to deal with halos using PI:

 

OK... I tried a different approach for reducing the halos. Here's what I did:

1. Cloned the image
2. Determined the left, right, top and bottom pixel values of the halo
3. Used pixel math to create a circle mask defined by the pixel boundaries. Here's the expression in case anyone wants it:
 

iif(sqrt((x()-(R+L)/2)^2 + (y()-(B+T)/2)^2) < (R-L)/2, 1, 0)

Also, in my case, the symbols are defined as: L=618, R=856, T=1200, B=1444. Those would change based on your image and the halo you were trying to remove.

4. Replaced the cloned image using that pixel math (dragged the blue triangle onto the cloned image from step 1)
5. Now, I have a mask. I used the Convolution process to blur the mask to taste.
6. Open up HistogramTransformation and bring the black point to the right until just before it clips. Also, I dragged the blue mid-tones to the right to get rid of the overall blue cast
7. Apply HistogramTransformation to image.

 

Before:

 

med_gallery_347158_14910_137828.jpg

 

After:

 

med_gallery_347158_14910_130716.jpg

 

The halo reduction was on epsilon CrA (HIP 93174) - the blue star - in the mid/upper left of the image.


Edited by jonnybravo0311, 14 February 2021 - 11:45 AM.


#13 imtl

imtl

    Fly Me to the Moon

  • *****
  • Moderators
  • Posts: 5,895
  • Joined: 07 Jun 2016
  • Loc: Down in a hole

Posted 14 February 2021 - 11:39 AM

Blasphemy!

 

I wrote this in another thread for a potential way to deal with halos using PI:

 

OK... I tried a different approach for reducing the halos. Here's what I did:

1. Cloned the image
2. Determined the left, right, top and bottom pixel values of the halo
3. Used pixel math to create a circle mask defined by the pixel boundaries. Here's the expression in case anyone wants it:
 

iif(sqrt((x()-(R+L)/2)^2 + (y()-(B+T)/2)^2) < (R-L)/2, 1, 0)

Also, in my case, the symbols are defined as: L=618, R=856, T=1200, B=1444. Those would change based on your image and the halo you were trying to remove.

4. Replaced the cloned image using that pixel math (dragged the blue triangle onto the cloned image from step 1)
5. Now, I have a mask. I used the Convolution process to blur the mask to taste.
6. Open up HistogramTransformation and bring the black point to the right until just before it clips. Also, I dragged the blue mid-tones to the right to get rid of the overall blue cast
7. Apply HistogramTransformation to image.

You could have saved yourself most of this by using GAME script


  • jonnybravo0311 likes this

#14 jonnybravo0311

jonnybravo0311

    Soyuz

  • *****
  • Moderators
  • Posts: 3,819
  • Joined: 05 Nov 2020
  • Loc: NJ, US

Posted 14 February 2021 - 11:47 AM

You could have saved yourself most of this by using GAME script

I could if GAME actually worked for me. Apparently it doesn't play nice with my Mac. Every time I try to use it, PI console just blows up with QT errors. Believe me, I WANT to use it LOL.

 

Edit: well... maybe upgrading to Big Sur and the latest version of PI solved the problem... I just went to grab a screenshot to show you how much it breaks... only to have it work just fine. I swear, it's not a lack of sleep this time! It really never did work for me in the past. That's why I had to come up with my crazy "create the mask manually" approach.


Edited by jonnybravo0311, 14 February 2021 - 11:51 AM.


#15 imtl

imtl

    Fly Me to the Moon

  • *****
  • Moderators
  • Posts: 5,895
  • Joined: 07 Jun 2016
  • Loc: Down in a hole

Posted 14 February 2021 - 11:49 AM

I could if GAME actually worked for me. Apparently it doesn't play nice with my Mac. Every time I try to use it, PI console just blows up with QT errors. Believe me, I WANT to use it LOL.

What's a Mac? rofl2.gif

 

** You should have a separate computer just for image processing shocked.gif


  • jonnybravo0311 likes this

#16 F.Meiresonne

F.Meiresonne

    Cosmos

  • *****
  • Posts: 8,504
  • Joined: 22 Dec 2003
  • Loc: Eeklo,Belgium

Posted 14 February 2021 - 12:11 PM

Norm, I think the image is darn good. Can you send me a link to the XISF final file? I want to try something out on this halo.

Dito....

 

only to small...I would boost it up...upload in gallery like png or something then post a link.

 

But it looks real good...for sure!



#17 F.Meiresonne

F.Meiresonne

    Cosmos

  • *****
  • Posts: 8,504
  • Joined: 22 Dec 2003
  • Loc: Eeklo,Belgium

Posted 14 February 2021 - 12:15 PM

 

 

 

Sorry I don't speak PI…

 

BQ

Neither  do I. Never understand what PI guys are talking about...shrug.gif



#18 ntph

ntph

    Viking 1

  • -----
  • topic starter
  • Posts: 743
  • Joined: 20 Oct 2012
  • Loc: Sudbury, ON

Posted 14 February 2021 - 12:46 PM

Wow--thanks, folks!

 

BQ:  great advice and explanations. Thanks for taking the time and effort. Now let's see if I actually understood and learn from it!

 

Jonny:  same as I did, as I explained using the method described by Charlie Bracken. And the same problem with GAME and my Mac running Catalina. Sometimes it crashes and sometimes it doesn't and when it doesn't, it doesn't always let me use a multipoint function. My workaround is to save everything I have open before I launch GAME; probably a more efficient way would be to save a project...

 

TxStars:  good point, as BQ notes my blues are clipped as well, so maybe deal with that at the same time that way, if I understand the physics and math of it...

 

Freddy:  https://astrob.in/4muylb/0/

 

Eyal:  My 10 year old MacBookPro beats the pants off my 4 year old HP laptop running PI. Neither is a high-end machine; I am just using what I have that workswink.gif  At least I never have to worry about the Mac shutting itself down to do an update in the middle of a session! The only reason I got a Windows machine was to be able to run AS and Registax native, because running Windows on a Mac just was too painful for me. I really don't care what platform I have to use, as long it just works the way it is supposed to! It has taken me literally years to get all my gear to work together, and still things hang or crash or just plain won't work unless you start everything all over from scratch (this mostly on the Windows side--I use the Windows machine outside and TeamViewer into my Mac inside where it is close to 40 C warmer! )



#19 imtl

imtl

    Fly Me to the Moon

  • *****
  • Moderators
  • Posts: 5,895
  • Joined: 07 Jun 2016
  • Loc: Down in a hole

Posted 14 February 2021 - 03:26 PM

There are several issues here. One is the "smaller" halo around Alnitak, the other, much much bigger faint glow stretching almost to the edge of the FOV. I'm still trying to get this right using GAME but I needed to stop for now. Don't think I've done a good job on it so far. :(

 

med_gallery_259665_321_9079274.png



#20 BQ Octantis

BQ Octantis

    Fly Me to the Moon

  • *****
  • Posts: 5,806
  • Joined: 29 Apr 2017
  • Loc: Mid Atlantic, USA

Posted 14 February 2021 - 06:37 PM

BQ:  great advice and explanations. Thanks for taking the time and effort. Now let's see if I actually understood and learn from it!

 

TxStars:  good point, as BQ notes my blues are clipped as well, so maybe deal with that at the same time that way, if I understand the physics and math of it...

I'm so sorry, Norm, but I started with one of your later images. When I inspected the original, I found the blue channel isn't clipped too badly:

 

gallery_273658_12412_429362.jpg

 

While what I said originally was true—Allnitak is brighter at lower wavelengths—the blue channel isn't blown out like in the later versions. This is very good news, as practically all of the data is usable.

 

Rather than continue down the PS babble, I'll advise this (which hopefully translates directly into PI babble): try starting with an R-RGB image. This uses the red channel for the luminosity while using RGB for the color. Here's an R-RGB from the original:

 

gallery_273658_12412_283910.png

Animated toggle

 

From there, you can feather in the luminosity from the original on areas where you like the detail. Like this:

 

gallery_273658_12412_103393.png

Animated toggle

 

Again, apologies for my confusion.

 

BQ


Edited by BQ Octantis, 15 February 2021 - 04:26 AM.


#21 ntph

ntph

    Viking 1

  • -----
  • topic starter
  • Posts: 743
  • Joined: 20 Oct 2012
  • Loc: Sudbury, ON

Posted 15 February 2021 - 06:12 PM

Certainly no need for apologies! Indeed, thank you for taking the time and effort to look at this, analyze the image and find information that leads to an approach to deal with the problem. It is easiest to see an obvious problem and try to find an easy and obvious way to deal with it, which was basically my approach: "brute force and bloody ignorance", as my grandfather used to say.. Your more analytical approach gets more to the root cause, which usually leads to a more robust solution. It is certainly more elegant. 

 

As far as translation from Ps to Pi, I don't think anything is lost. (I love your toggled animations, by the way.) I am assuming that the step from the R-RGB to the final version in its simplest form involves appropriate masks and intensity transformations? In another Ps-Pi comparison, this explanation by BQ Octantis has similarities to the way a Pi guru named Adam Block approaches problems; sort of a first principles approach. Thanks for your advice. 

 

Clear skies and stay well--hopefully the end of current troubles is nearing for all of us. 



#22 BQ Octantis

BQ Octantis

    Fly Me to the Moon

  • *****
  • Posts: 5,806
  • Joined: 29 Apr 2017
  • Loc: Mid Atlantic, USA

Posted 16 February 2021 - 02:38 AM

I am assuming that the step from the R-RGB to the final version in its simplest form involves appropriate masks and intensity transformations?

I just pasted in the original, applied a black mask, and then used a soft, ~50% flow brush to feather in the highlights on the periphery. I made no intensity changes, but I think I used Lighten as the layer mode—which would have changed the contrast of the details, especially in the lower right. If I were being more careful, I'd probably examine the finer details to feather in in other areas—for example, within the horse's head itself are a few puffs of dust that are absent in the R-RGB.

 

Fingers crossed on the end of troubles! fingerscrossed.gif

 

BQ



#23 MikiSJ

MikiSJ

    Surveyor 1

  • *****
  • Posts: 1,890
  • Joined: 24 Feb 2006
  • Loc: San Jose, CA

Posted 16 February 2021 - 03:48 AM

Just a thought.

 

I was playing around with 'Closed Loop Slew' in TheSKYX and one of the targets I entered was the Trapezium. Since the image duration was very short (<10s) I got a really nice image of the major stars plus some of the lesser seen stars in the 'Trap' and I started to think about imaging M42 and rather than allow the 'Trap' to burn into the image, why not mask out the 'Trap' in the longer M42 image and lay in the shorter view of the 'Trap' stars?

 

Couldn't something similar be done with Alnitak? The Alnitak image could be feathered into the larger image using one of the brushes in Photoshop.



#24 imtl

imtl

    Fly Me to the Moon

  • *****
  • Moderators
  • Posts: 5,895
  • Joined: 07 Jun 2016
  • Loc: Down in a hole

Posted 16 February 2021 - 05:48 AM

Just a thought.

 

I was playing around with 'Closed Loop Slew' in TheSKYX and one of the targets I entered was the Trapezium. Since the image duration was very short (<10s) I got a really nice image of the major stars plus some of the lesser seen stars in the 'Trap' and I started to think about imaging M42 and rather than allow the 'Trap' to burn into the image, why not mask out the 'Trap' in the longer M42 image and lay in the shorter view of the 'Trap' stars?

 

Couldn't something similar be done with Alnitak? The Alnitak image could be feathered into the larger image using one of the brushes in Photoshop.

You can use HDR composition on whatever you like. I personally would not go for all that trouble just for one bright star that can be dealt with in postprocessing and with filters that do not give bad halos.

 

The trapezium is star structure that is in the heart of the nebula. Alnitak is just a star (okay a monster). 

But sure, you could do something like that, I don't see why not.



#25 SeymoreStars

SeymoreStars

    Soyuz

  • *****
  • Posts: 3,614
  • Joined: 08 May 2014
  • Loc: Pennsyltucky

Posted 16 February 2021 - 01:47 PM

Use a 3nm Ha filter for Lum exposures and short exposures of RGB.

 

 

 

get.jpg?insecure


Edited by SeymoreStars, 16 February 2021 - 01:48 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