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 Question for PI Users

  • Please log in to reply
14 replies to this topic

#1 sdeming

sdeming

    Ranger 4

  • -----
  • topic starter
  • Posts: 359
  • Joined: 29 May 2020
  • Loc: Charlotte, North Carolina - USA

Posted 12 March 2021 - 08:12 AM

Hello Friends!

 

I started using PixInsight a few months ago and really like it. It's quite robust. My question is in regards to stretching. I have always used either auto stretch or EZ soft stretch as my initial stretching tool. However, I recently read in the PixInsight manual that using the ArcsinhStretch is the best way to maintain original color. So, I have been using the ArcsinhStretch as my initial stretching tool on my last few images, after color calibration, background extraction, decon and denoise. I'm not sure if this is how I should be doing my initial stretch. Although it does seem to maintain more of the color, it also seems very saturated. Even the stars are saturated.

 

Below are two different processes of Rosette I took in February. The first is using auto stretch and the second is using ArcsinhStretch. I actually had to unsaturate the second image a bit. Any thoughts or recommendations would be greatly appreciated.

 

Thank you!

 

Scott

Attached Thumbnails

  • Rosette L JPEG resized.jpg
  • Rosette reprocessed 3 11 21 JPEG resized.jpg

Edited by sdeming, 12 March 2021 - 08:15 AM.

  • kathyastro likes this

#2 cybermayberry

cybermayberry

    Vostok 1

  • -----
  • Posts: 199
  • Joined: 26 Aug 2010

Posted 12 March 2021 - 12:26 PM

Unless you profiling your monitors with color calibration equipment you really have no way to judge the accuracy of the colors displayed.

I am no PI expert by any means, but the fact they have multiple way to stretch leads me to concluded their is no "universal best way". I often clone the image an try several way to see which I think looks best to me.

I don't worry to much about the color balance or saturation at that point. I adjust those later anyway to my liking.

In the end it's pretty subjective anyway.

Nice rosette, great detail



#3 sdeming

sdeming

    Ranger 4

  • -----
  • topic starter
  • Posts: 359
  • Joined: 29 May 2020
  • Loc: Charlotte, North Carolina - USA

Posted 12 March 2021 - 12:54 PM

Unless you profiling your monitors with color calibration equipment you really have no way to judge the accuracy of the colors displayed.

I am no PI expert by any means, but the fact they have multiple way to stretch leads me to concluded their is no "universal best way". I often clone the image an try several way to see which I think looks best to me.

I don't worry to much about the color balance or saturation at that point. I adjust those later anyway to my liking.

In the end it's pretty subjective anyway.

Nice rosette, great detail

Thank you for your feedback and the kind words. Much appreciated!



#4 Peregrinatum

Peregrinatum

    Surveyor 1

  • *****
  • Posts: 1,736
  • Joined: 27 Dec 2018
  • Loc: South Central Valley, Ca

Posted 12 March 2021 - 01:45 PM

I have found that using a MaskedStretch is the best way to preserve star color.



#5 DEnc

DEnc

    Sputnik

  • -----
  • Posts: 33
  • Joined: 19 Feb 2016

Posted 12 March 2021 - 02:04 PM

Christopher Foster has posted a link to his PixInsight workflow and although its title specifies CMOS cameras, I find it very helpful in processing my CCD images.  

 

He describes ArcSinHStretch and its pitfalls, and how he incorporates it into stretching.

 

https://drive.google...Mp4gwA9zIs4FUnw


  • sdeming likes this

#6 Stelios

Stelios

    Voyager 1

  • *****
  • Posts: 11,110
  • Joined: 04 Oct 2003
  • Loc: West Hills, CA

Posted 12 March 2021 - 02:30 PM

I have used MaskedStretch, ArcSinhstretch, Curves only, and STF->Histogram.

 

I personally detest MaskedStretch--it does little or nothing for me, muted city. Have yet to find one image where it got me even close enough to adjust with curves. It seems best suited for those who like the "visual" look, objects close to how they appear through the eyepiece. I prefer bolder images.

ArcSinhstretch works OK on some objects, not so much on others. I don't like that it doesn't leave me any margin to raise the black point. Perhaps I use it poorly.

Curves is great for difficult cases, as you can stop, review, and mask as needed, but it's slower. Might as well just use Photoshop.

 

So, overall, I use the much maligned STF->Histogram. I then usually have to take the result down a peg using Curves, or (in the case of Galaxies) HDRMT. As long as you do that, it works *very* much to my liking.

 

As for color, I use PCC for all broadband images. I always separate the L channel from color and process L for sharpness, C for color noise, then stretch both, denoise color once more, and recombine. The trick to preserving color is to make sure that the L channel's brightest areas are less than 0.8 (on a 0-1 scale). Of course you can't do that for saturated star and galaxy cores, but surrounding detail should be kept (whether by stretching or HDRMT) to 0.8 or less. 


  • dswtan, TrustyChords, limeyx and 1 other like this

#7 mewmartigan

mewmartigan

    Apollo

  • *****
  • Posts: 1,325
  • Joined: 02 Jul 2008
  • Loc: outside of Pittsburgh, PA

Posted 12 March 2021 - 02:35 PM

I will usually either used MaskedStretch or ArcSinh stretch. I won't push them too far though. Then I will use masks to further highlight or saturate particular regions. This helps the stars from getting to nuts.

 

I have not really had any luck removing stars with Starnet and trying to add them back in later. It always seems to leave artifacts around and can take forever to clone stamp out.


  • limeyx likes this

#8 TrustyChords

TrustyChords

    Mariner 2

  • *****
  • Posts: 290
  • Joined: 26 Jun 2017

Posted 12 March 2021 - 02:57 PM

Something I learned from someone else here, I use ArcSinh stretch and Histogram Stretch together.

 

I will usually ArcSinh stetch until I start to see color decently but the background is still pretty dark, and then finish with a Histogram Stretch.

 

You can actually have the ArcSinh stretch preview open and STF preview open at the same time and use them together to preview the combination at once-- then apply ArcSinh first, copy STF to Histogram and apply that. This way is nice for black point too, since you shouldn't be clipping any, and then you can raise the black point up to the baseline where you want it in Histogram stretch.



#9 sdeming

sdeming

    Ranger 4

  • -----
  • topic starter
  • Posts: 359
  • Joined: 29 May 2020
  • Loc: Charlotte, North Carolina - USA

Posted 12 March 2021 - 03:22 PM

I have found that using a MaskedStretch is the best way to preserve star color.

Do you stretch the stars? I know how to do star masks, but I use that to either saturate the stars, control the brightness or make them smaller. Is there a way to mask the stars and have them not effected by the stretch?



#10 sdeming

sdeming

    Ranger 4

  • -----
  • topic starter
  • Posts: 359
  • Joined: 29 May 2020
  • Loc: Charlotte, North Carolina - USA

Posted 12 March 2021 - 03:25 PM

I have used MaskedStretch, ArcSinhstretch, Curves only, and STF->Histogram.

 

I personally detest MaskedStretch--it does little or nothing for me, muted city. Have yet to find one image where it got me even close enough to adjust with curves. It seems best suited for those who like the "visual" look, objects close to how they appear through the eyepiece. I prefer bolder images.

ArcSinhstretch works OK on some objects, not so much on others. I don't like that it doesn't leave me any margin to raise the black point. Perhaps I use it poorly.

Curves is great for difficult cases, as you can stop, review, and mask as needed, but it's slower. Might as well just use Photoshop.

 

So, overall, I use the much maligned STF->Histogram. I then usually have to take the result down a peg using Curves, or (in the case of Galaxies) HDRMT. As long as you do that, it works *very* much to my liking.

 

As for color, I use PCC for all broadband images. I always separate the L channel from color and process L for sharpness, C for color noise, then stretch both, denoise color once more, and recombine. The trick to preserving color is to make sure that the L channel's brightest areas are less than 0.8 (on a 0-1 scale). Of course you can't do that for saturated star and galaxy cores, but surrounding detail should be kept (whether by stretching or HDRMT) to 0.8 or less. 

Thank you! Where do I find PCC and what is it?



#11 sdeming

sdeming

    Ranger 4

  • -----
  • topic starter
  • Posts: 359
  • Joined: 29 May 2020
  • Loc: Charlotte, North Carolina - USA

Posted 12 March 2021 - 03:26 PM

Something I learned from someone else here, I use ArcSinh stretch and Histogram Stretch together.

 

I will usually ArcSinh stetch until I start to see color decently but the background is still pretty dark, and then finish with a Histogram Stretch.

 

You can actually have the ArcSinh stretch preview open and STF preview open at the same time and use them together to preview the combination at once-- then apply ArcSinh first, copy STF to Histogram and apply that. This way is nice for black point too, since you shouldn't be clipping any, and then you can raise the black point up to the baseline where you want it in Histogram stretch.

Nice! I'm gonna try this! Thank you!



#12 sdeming

sdeming

    Ranger 4

  • -----
  • topic starter
  • Posts: 359
  • Joined: 29 May 2020
  • Loc: Charlotte, North Carolina - USA

Posted 12 March 2021 - 03:39 PM

Christopher Foster has posted a link to his PixInsight workflow and although its title specifies CMOS cameras, I find it very helpful in processing my CCD images.  

 

He describes ArcSinHStretch and its pitfalls, and how he incorporates it into stretching.

 

https://drive.google...Mp4gwA9zIs4FUnw

Thank you! I will certainly read this.



#13 sdeming

sdeming

    Ranger 4

  • -----
  • topic starter
  • Posts: 359
  • Joined: 29 May 2020
  • Loc: Charlotte, North Carolina - USA

Posted 12 March 2021 - 03:53 PM

I have used MaskedStretch, ArcSinhstretch, Curves only, and STF->Histogram.

 

I personally detest MaskedStretch--it does little or nothing for me, muted city. Have yet to find one image where it got me even close enough to adjust with curves. It seems best suited for those who like the "visual" look, objects close to how they appear through the eyepiece. I prefer bolder images.

ArcSinhstretch works OK on some objects, not so much on others. I don't like that it doesn't leave me any margin to raise the black point. Perhaps I use it poorly.

Curves is great for difficult cases, as you can stop, review, and mask as needed, but it's slower. Might as well just use Photoshop.

 

So, overall, I use the much maligned STF->Histogram. I then usually have to take the result down a peg using Curves, or (in the case of Galaxies) HDRMT. As long as you do that, it works *very* much to my liking.

 

As for color, I use PCC for all broadband images. I always separate the L channel from color and process L for sharpness, C for color noise, then stretch both, denoise color once more, and recombine. The trick to preserving color is to make sure that the L channel's brightest areas are less than 0.8 (on a 0-1 scale). Of course you can't do that for saturated star and galaxy cores, but surrounding detail should be kept (whether by stretching or HDRMT) to 0.8 or less. 

Another question - do you do the stretching after all of the other steps, such as color calibration, denoise, decon, background extraction, etc.?



#14 Stelios

Stelios

    Voyager 1

  • *****
  • Posts: 11,110
  • Joined: 04 Oct 2003
  • Loc: West Hills, CA

Posted 12 March 2021 - 06:47 PM

Another question - do you do the stretching after all of the other steps, such as color calibration, denoise, decon, background extraction, etc.?

PCC = Photometric Color Calibration

 

Before stretching: Dynamic Crop, DBE, Deconvolution (on L only), Denoise (mild). 

 

After stretching: depends, but usually HDRMT (on L), TGVDenoise (on C), LRGB combine, and then LHE (Local Histogram Equalization), ET (exponential transformation), CS (color saturation), MT (morphological transformation for star reduction) and touches as needed of more TGVDenoise, HT, Curves, etc. Not all are needed every time, of course!

 

Other tools that can have dramatic effect are MureDenoise (must be done before anything else, right after stacking, but careful with DBE after--don't recommend if you have huge gradients) and LHH (LocalHistogramHyperbolization). 

 

I often use the EZProcessingSuite scripts (especially Denoise and star reduction) in place of the above. But I recommend learning how to do the processing oneself before relying on it, as otherwise you can't really tell if it's working well or not. 

 

And whether you have any current use for it or not, I suggest getting familiar with the *fantastic* GAME script for creating masks. It really is a game-changer. 


  • dswtan, SnowWolf, limeyx and 2 others like this

#15 sdeming

sdeming

    Ranger 4

  • -----
  • topic starter
  • Posts: 359
  • Joined: 29 May 2020
  • Loc: Charlotte, North Carolina - USA

Posted 13 March 2021 - 02:13 PM

PCC = Photometric Color Calibration

 

Before stretching: Dynamic Crop, DBE, Deconvolution (on L only), Denoise (mild). 

 

After stretching: depends, but usually HDRMT (on L), TGVDenoise (on C), LRGB combine, and then LHE (Local Histogram Equalization), ET (exponential transformation), CS (color saturation), MT (morphological transformation for star reduction) and touches as needed of more TGVDenoise, HT, Curves, etc. Not all are needed every time, of course!

 

Other tools that can have dramatic effect are MureDenoise (must be done before anything else, right after stacking, but careful with DBE after--don't recommend if you have huge gradients) and LHH (LocalHistogramHyperbolization). 

 

I often use the EZProcessingSuite scripts (especially Denoise and star reduction) in place of the above. But I recommend learning how to do the processing oneself before relying on it, as otherwise you can't really tell if it's working well or not. 

 

And whether you have any current use for it or not, I suggest getting familiar with the *fantastic* GAME script for creating masks. It really is a game-changer. 

Stelio, thank you so much for the awesome workflow!




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