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

Bloated RGB stars combining with Ha?

This topic has been archived. This means that you cannot reply to this topic.
42 replies to this topic

#1 Footbag

Footbag

    Cosmos

  • *****
  • topic starter
  • Posts: 9115
  • Joined: 13 Apr 2009

Posted 17 June 2015 - 09:16 AM

Finally a clear night!  Last night I got some RGB to finish up an image of Sadr that I've been working on. 

 

I'm processing it and the stars in my RGB channel are bloated.  Looks like clouds or conditions, but whatever it is, it's there.  Problem is, when I combine my Ha with the RGB, I get halos around the stars.  What would you guys do? 

 

How do I get them to line up nicely?



#2 anismo

anismo

    Fly Me to the Moon

  • *****
  • Posts: 5932
  • Joined: 30 Jan 2014

Posted 17 June 2015 - 09:19 AM

How about MT of the RGB stars before the channel combination?

 

Or carboni's halo reduction action once it is all combined?



#3 Madratter

Madratter

    Voyager 1

  • *****
  • Posts: 10796
  • Joined: 14 Jan 2013

Posted 17 June 2015 - 09:22 AM

That is a bear and I don't have a great answer. If just replacing the stars, you can stretch the RGB with the stars less. But that requires you to do a starless version.

#4 jrcrilly

jrcrilly

    Refractor wienie no more

  • *****
  • Administrators
  • Posts: 35465
  • Joined: 30 Apr 2003

Posted 17 June 2015 - 09:24 AM

Achieving the same star sizes with RGB as with Ha requires increasing the Ha exposure time by a large factor. Balancing the star sizes can be achieved by either selecting exposure times that provide similar FWHM (most desirable) or by deconvolving the RGB frame until the FWHM is similar (less desirable). Or by using post-processing trickery (least desirable). 



#5 Footbag

Footbag

    Cosmos

  • *****
  • topic starter
  • Posts: 9115
  • Joined: 13 Apr 2009

Posted 17 June 2015 - 09:26 AM

Well, you have given me good options.  I think I'll have to try them both.  MT hadn't worked for me in the past, but it's still the first thing I thought. 

 

A starless image is the other I was considering.  It' going to be tough to prevent that overly red look.  We will see.

 

I can also give decon a look, but I think I'll have more control in MT. 



#6 josh smith

josh smith

    Fly Me to the Moon

  • *****
  • Posts: 5332
  • Joined: 13 Oct 2013

Posted 17 June 2015 - 09:39 AM

Definitely the best results would come from a starless version, but it will be more work.  If you can completely remove the stars and just put the rgb in, there's no worry of mismatch.



#7 David Ault

David Ault

    Gemini

  • -----
  • Posts: 3424
  • Joined: 25 Sep 2010

Posted 17 June 2015 - 10:03 AM

There are a few methods for doing this.  Most of they rely on taking the stars from the broadband data (typically the red channel) and putting in the Ha data.  I particularly like this method.  Josh's idea of doing starless tone-mapping will probably get you the best results though since you can play with the stars separately.

 

EDIT:

I'll add some more comments for the method in the link I provided, since that page is lacking many details on how to actually do everything.  I didn't find it necessary to make multiple passes of continuum maps and pseudo red images if I got a good pass of noise reduction on the continuum map.  To do that I used the Ha channel permanently stretched with STF auto stretch settings for local support in TGVDenoise (with default settings for shadows highlights and midtones).  When generating the continuum map you also need to turn on the rescale option for PixelMath otherwise you'll get clipped data.  It's either that or manually scale the data, for example C = B/N becomes C = m*B/N + b and B = N*C becomes B = N*(C-b)/m.  You need to pick values for m & b such that no pixels are clipped, which you can check with the Statistics process.  For the last data set I used I chose m=0.1 & b=0.1.  The nice thing about this change is that the resulting image has the same intensity profile as the original, where as if you just rescale the data then you'll have to redo some kind of linear fit.

 

In the past I had used something like this: a*max(Ha, R) + ~a*((Ha+R)/2), where a is the blending factor between the average and max functions (this assumes a LinearFit was run prior so that the intensities of the images match up).  It works reasonably well but you get some of the noise from the red channel showing up in weird ways due to the max function.

 

Regards,

David


Edited by David Ault, 17 June 2015 - 10:29 AM.


#8 Footbag

Footbag

    Cosmos

  • *****
  • topic starter
  • Posts: 9115
  • Joined: 13 Apr 2009

Posted 17 June 2015 - 10:21 AM

Thanks Josh and David.  I think I've been dancing around the solution, but I hope to get it processed tonight.

 

A friend just texted to see if I got some shooting in before the clouds came in.  I didn't know it got cloudy, and I didn't scrutinize the stacks yet. 

 

David, I"ll be reading and rereading that a bunch of times tonight. ;)


Edited by Footbag, 17 June 2015 - 10:52 AM.


#9 Footbag

Footbag

    Cosmos

  • *****
  • topic starter
  • Posts: 9115
  • Joined: 13 Apr 2009

Posted 17 June 2015 - 04:48 PM

Stuck on step 1...

R/Ha is giving me a blank image in Pixelmath. What am I doing wrong?

Or is it clipping and those are the numbers I need to select. That must be it. Darn, my daughter just took the PC.

Edited by Footbag, 17 June 2015 - 04:53 PM.


#10 David Ault

David Ault

    Gemini

  • -----
  • Posts: 3424
  • Joined: 25 Sep 2010

Posted 17 June 2015 - 05:07 PM

It's probably clipping.  In general you are dividing a broadband image which will be exposed to higher ADU values by a narrowband image which will barely get above the read noise, so you are likely to get values above 1.0 which will be clipped.

 

You can play with the scaling values (m and b) but if you just want a quick pass you might just turn on the rescale option which will normalize the output.  The way I picked the scaling values was by running PixelMath and dropping the value of m until the maximum pixel value was below 1.0, (keeping be at 0.0).  Then I increased b until the minimum value was above 0.0, then decreased m again.  I'm sure you could pre-compute the m and b values using the min and max of both images, but I just found it easier to do a couple passes of PixelMath.

 

Regards,

David



#11 Footbag

Footbag

    Cosmos

  • *****
  • topic starter
  • Posts: 9115
  • Joined: 13 Apr 2009

Posted 17 June 2015 - 05:16 PM

Thanks David!

 

I did have rescale on.  That's what gives me a completely black image.  When it's turned off, I think it's completely white.  I'm sure I can figure it out, just dind't know why it was black when rescaled with no upper or lower limit. 



#12 Peter in Reno

Peter in Reno

    Voyager 1

  • *****
  • Posts: 10415
  • Joined: 15 Jul 2008

Posted 17 June 2015 - 06:26 PM

Have you tried PI's NBRGBCombination script under Utilities? It's based on Vicent Peris and Harry's Add Ha to a Galaxy tutorials.

 

Peter



#13 David Ault

David Ault

    Gemini

  • -----
  • Posts: 3424
  • Joined: 25 Sep 2010

Posted 17 June 2015 - 09:19 PM

I'm really not sure why you get all black or all white.  Here's my PixelMath setup for either method:

PseudoHaR_1.jpg

With manual scaling

 

PseudoHaR_2.jpg

With automatic scaling

 

Peter, I've tried the NBRGBCombination script and it does a good job of keeping the RGB stars but looses a lot of the detail from the narrowband data.  The multi-channel synthesis scripts don't handle the stars well, at least not with my 3nm H-alpha data.  This other method merges the stars well and gets more of the narrowband detail in the final image.

 

Regards,

David



#14 schmeah

schmeah

    Fly Me to the Moon

  • *****
  • Posts: 5164
  • Joined: 26 Jul 2005

Posted 17 June 2015 - 10:29 PM

It's easy to do in Photoshop. A simple lighten blend of an HaR combo which maintains the red star size, which is then used as the red channel in an RGB combine.

 

Derek



#15 David Ault

David Ault

    Gemini

  • -----
  • Posts: 3424
  • Joined: 25 Sep 2010

Posted 17 June 2015 - 11:23 PM

Derek,

 

The lighten blend mode is a maximum function.  The opacity slider does an alpha blend (also called linear interpolation, linear blend, alpha compositing, etc.) between the resulting function and the top layer which is almost identical to the equation I posted above in this section:

In the past I had used something like this: a*max(Ha, R) + ~a*((Ha+R)/2), where a is the blending factor between the average and max functions (this assumes a LinearFit was run prior so that the intensities of the images match up).  It works reasonably well but you get some of the noise from the red channel showing up in weird ways due to the max function.

 

There is a difference in this equation in that some of the original Ha signal is blended directly in, but it's the same idea.  In your example, you would replace (Ha+R)/2 with just R.  You can see that when a=1 you get only the max(Ha, R) result which equates to an opacity of 0 (assuming the R data is the top layer).  At a=1 you get only the R result or an effective opacity of 100.

 

The only reason I brought this up is so that people can see the math behind some of what Photoshop does and how it equates to what you can do in programs like PixInsight.

 

Regards,

David



#16 Footbag

Footbag

    Cosmos

  • *****
  • topic starter
  • Posts: 9115
  • Joined: 13 Apr 2009

Posted 18 June 2015 - 07:07 AM

David,

Am I supposed to be creating a continuum map with this or is it creating a full blended image? Ha with R stars?

I seem to be getting the latter. Even used it in my final. But I'm expecting a nebulaless starfield, no?

#17 shawnhar

shawnhar

    Cosmos

  • *****
  • Posts: 9393
  • Joined: 25 Jun 2010

Posted 18 June 2015 - 07:25 AM

Finally a clear night!  Last night I got some RGB to finish up an image of Sadr that I've been working on. 

 

I'm processing it and the stars in my RGB channel are bloated.  Looks like clouds or conditions, but whatever it is, it's there.  Problem is, when I combine my Ha with the RGB, I get halos around the stars.  What would you guys do? 

 

How do I get them to line up nicely?

 

 Well I'm a total PI newb so take it for what it's worth, but I created a star mask for the rgb and used the morph erosion, (I guess that's MT?) a few times, blured the star mask pretty good, then used the LRGB tool to combine and the halos were almost completely gone.



#18 Footbag

Footbag

    Cosmos

  • *****
  • topic starter
  • Posts: 9115
  • Joined: 13 Apr 2009

Posted 18 June 2015 - 08:06 AM

 

Finally a clear night!  Last night I got some RGB to finish up an image of Sadr that I've been working on. 

 

I'm processing it and the stars in my RGB channel are bloated.  Looks like clouds or conditions, but whatever it is, it's there.  Problem is, when I combine my Ha with the RGB, I get halos around the stars.  What would you guys do? 

 

How do I get them to line up nicely?

 

 Well I'm a total PI newb so take it for what it's worth, but I created a star mask for the rgb and used the morph erosion, (I guess that's MT?) a few times, blured the star mask pretty good, then used the LRGB tool to combine and the halos were almost completely gone.

 

 

 

I did that in my first run and it turned out very nicely, but my color balance was off.  I hope to give this a few different tries with each method, but it's all about getting PC time from the kids.  And I'm pretty dead tired when they go to bed. 

 

It's easier to make time to image. 



#19 David Ault

David Ault

    Gemini

  • -----
  • Posts: 3424
  • Joined: 25 Sep 2010

Posted 18 June 2015 - 08:14 AM

So the steps are:

1) Create the continuum map using: R/Ha with rescale on or m*R/Ha+b

- It should like primarily like the stars probably with a dimple in the center of the brighter stars

- You may also see hints of the nebulosity where the red and Ha values differ

2) Denoise the continuum map

- You want to remove as much noise as possible without destroying any of the structures

3) Make a new 'broadband' image from the continuum map and the Ha image: C*Ha or N*(C-b)/m

- If you used the rescale option you will need to do a LinearFit to the R data to get it back to the same intensity profile

 

Regards,

David



#20 David Ault

David Ault

    Gemini

  • -----
  • Posts: 3424
  • Joined: 25 Sep 2010

Posted 18 June 2015 - 08:17 AM

Shawn,

 

What, you don't want to solve everything with math now that you have PixInsight!?  Buzzkill ;)

 

Regards,

David



#21 karambit27

karambit27

    Viking 1

  • *****
  • Posts: 880
  • Joined: 15 Aug 2013

Posted 09 July 2015 - 10:04 AM

There are a few methods for doing this.  Most of they rely on taking the stars from the broadband data (typically the red channel) and putting in the Ha data.  I particularly like this method.  Josh's idea of doing starless tone-mapping will probably get you the best results though since you can play with the stars separately.

 

EDIT:

I'll add some more comments for the method in the link I provided, since that page is lacking many details on how to actually do everything.  I didn't find it necessary to make multiple passes of continuum maps and pseudo red images if I got a good pass of noise reduction on the continuum map.  To do that I used the Ha channel permanently stretched with STF auto stretch settings for local support in TGVDenoise (with default settings for shadows highlights and midtones).  When generating the continuum map you also need to turn on the rescale option for PixelMath otherwise you'll get clipped data.  It's either that or manually scale the data, for example C = B/N becomes C = m*B/N + b and B = N*C becomes B = N*(C-b)/m.  You need to pick values for m & b such that no pixels are clipped, which you can check with the Statistics process.  For the last data set I used I chose m=0.1 & b=0.1.  The nice thing about this change is that the resulting image has the same intensity profile as the original, where as if you just rescale the data then you'll have to redo some kind of linear fit.

 

In the past I had used something like this: a*max(Ha, R) + ~a*((Ha+R)/2), where a is the blending factor between the average and max functions (this assumes a LinearFit was run prior so that the intensities of the images match up).  It works reasonably well but you get some of the noise from the red channel showing up in weird ways due to the max function.

 

Regards,

David

David I have been trying this method with decent results but I am having trouble with the last Ha boosting part. How do you transfer the CIE a* and CIE b* components over?



#22 David Ault

David Ault

    Gemini

  • -----
  • Posts: 3424
  • Joined: 25 Sep 2010

Posted 09 July 2015 - 12:54 PM

You can transfer the CIE a* & b* channels using the ChannelExtraction and ChannelCombination processes.  Use ChannelExtraction on both the image you want the luminance from and the one you want the chrominance from (a* & b*).  Using ChannelCombination use the L data from the image you want as the luminance and use the a* & b* channels from the other image.

 

Let me know if that answered your question.

 

Regards,

David



#23 karambit27

karambit27

    Viking 1

  • *****
  • Posts: 880
  • Joined: 15 Aug 2013

Posted 09 July 2015 - 01:09 PM

Ok that is how I thought it should be done but When I do that i don't get anything like what I'd expect. The combined imaged is basically black with a couple pinpoint stars. I'll try and provide some examples once I get home from work. Is this supposed to be done while the  data is linear or non-linear?



#24 David Ault

David Ault

    Gemini

  • -----
  • Posts: 3424
  • Joined: 25 Sep 2010

Posted 09 July 2015 - 01:46 PM

The channel extraction/combination processes should work regardless if your data is linear or non-linear.  I've only used their Ha boosting process once before and I did it when the data was linear.  I didn't like the results and ended up going a different route.

 

Some examples would definitely be helpful.

 

Regards,

David



#25 karambit27

karambit27

    Viking 1

  • *****
  • Posts: 880
  • Joined: 15 Aug 2013

Posted 09 July 2015 - 05:42 PM

Ok so here is what I am seeing. On the Left is the HaRGB image top then the CIE a* and CIE b* below it. On the left is the HaGB image top then the CIE L*  and the resulting combination of the CIE L , a, and b below it. All of them have a auto STF applied. I can't figure out why the result is the way it is. Appreciate your help David.

Attached Thumbnails

  • Capture.JPG

Edited by karambit27, 09 July 2015 - 05:43 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