Pixinsight improved image stretching

[geeklee]

4 hours ago, Scooot said:

That’s a fantastic image. 

2 hours ago, Gunshy said:

That's spectacular! 

Thanks both for the feedback! 😊

2 hours ago, Gunshy said:

Any tips, tricks, or difficulties you encountered along the way to producing this would be greatly appreciated.

At the moment, not really Dave.  Your continued notes in this thread help a lot (and I still need to spend time with the documentation).  It was getting a feel for where to focus the stretch in later iterations.  As per my previous messages, I'm still not all there with contrast but getting more comfortable in the tool itself.

[scotty38]

Haven't really had chance for a further play with this but I'd been struggling with "placing the point on the curve" etc etc but the video cleared that up so I now know how to do it and the numbers to look for...

I can now get images stretching by my actions rather than "click and hope"...

Despite saying it already, I will keep trying.

[mike1485]

22 hours ago, geeklee said:

My first proper go at stretching three masters - Ha, OIII and SII - using the GHS script with multiple iterations.  It controlled Eta and Mu Gem well during the stretch.

https://www.astrobin.com/swc0q9/

Thanks again @mike1485 and @Gunshy for their efforts.  A useful addition to the toolbox.

That really is an awesome image - very well done!

[Annehouw]

I had a go at it for an image that I still hadn't processed. To be honest, it took me a while to get a feeling of the interplay of the different parameters, but the video helped tremendously. I used GHS for the basic stretch and did a bunch of global and local stuff afterwards to bring out color and dust lanes (there's the low contrast "three blade propellor" -two o'clock position close to the core-, but many others as well).  I will chew on this rendition for a while to think whether and how much it is overprocessed or not, but I would like to say thank you to Dave and Mike for adding another tool to the toolbox!

 

M13:

 

Edited January 4, 2022 by Annehouw

[scotty38]

Wow that's some image!

[Annehouw]

To give a better idea of what GHS can accomplish, I made a compilation of 4 stretching methods:

1. Generalised Hyperbolic Stretch (GHS),
2. Masked Stretch (MS),
3. Histogram Transformation (HT)
4. Arcshinh stretch (ASH), which is actually a combination of a very mild HT stretch followed by ASH (otherwise it becomes a bit ugly...).

The stretches are not 100% equivalent, but it gives an idea.

 

The inserts show the bright yellow star on the right and a star to the left of this star. 

As can be seen, GHS looks a lot like MS. MS retains a bit more color, but the downside of MS is that it has created a harsh plateau in the star brightness, resulting in a distracting dark ring in the bright star. 

HT puts a lot of light into the halo and gives a relatively contrasty image (to be corrected by Local Histogram Transormation or HDRMT).

ASH is the colorful one. That is nice. Unfortunately, it also gives rise the the scary phenomenon of "rainbow stars", which you can see in the inset of the smaller star. This is an artifact that is very difficult to correct.

 

For the "final" M13 image I used the GHS stretch as a basis followed by stretches to bring a bit of contrast into the image, increased saturation on the globular cluster and a bit on the stars, desaturation on the background and local stretches to bring out the deteils in the core of M13.

 

 

[nephilim]

On 04/01/2022 at 15:08, Annehouw said:

I had a go at it for an image that I still hadn't processed. To be honest, it took me a while to get a feeling of the interplay of the different parameters, but the video helped tremendously. I used GHS for the basic stretch and did a bunch of global and local stuff afterwards to bring out color and dust lanes (there's the low contrast "three blade propellor" -two o'clock position close to the core-, but many others as well).  I will chew on this rendition for a while to think whether and how much it is overprocessed or not, but I would like to say thank you to Dave and Mike for adding another tool to the toolbox!

 

M13:

 

@Annehouw That is stunning. 

[Richard_]

Nice work @Annehouw, I've been meaning to try this approach but haven't had a chance to do it yet. 

Question: has anyone noticed if GHS improves subsequent StarNet process? With my current approaches (HT and ArcsinhStretch + HT) StarNet hasn't always caught all stars and leaves some mottling behind and around the edges of stars. It's less noticeable with Arcsinhstretch but it's still there, curious to know if GHS improves this further or not?

[Annehouw]

@Richard_

I do not use Starnet anymore, but StarXterminator instead.

I did try StarXterminator after GHS, but it left white dots everywhere where stars were removed. I think that the GHS starshapes deviate too much to the starshapes on which the model has been trained.

Best to use a mild HT for that purpose.

(StarXterminator als works on linear images in PI, but the results are slightly inferior compared to removal after a mild HT).

[Gunshy]

30 minutes ago, Annehouw said:

To give a better idea of what GHS can accomplish, I made a compilation of 4 stretching methods:

1. Generalised Hyperbolic Stretch (GHS),
2. Masked Stretch (MS),
3. Histogram Transformation (HT)
4. Arcshinh stretch (ASH), which is actually a combination of a very mild HT stretch followed by ASH (otherwise it becomes a bit ugly...).

The stretches are not 100% equivalent, but it gives an idea.

 

The inserts show the bright yellow star on the right and a star to the left of this star. 

As can be seen, GHS looks a lot like MS. MS retains a bit more color, but the downside of MS is that it has created a harsh plateau in the star brightness, resulting in a distracting dark ring in the bright star. 

HT puts a lot of light into the halo and gives a relatively contrasty image (to be corrected by Local Histogram Transormation or HDRMT).

ASH is the colorful one. That is nice. Unfortunately, it also gives rise the the scary phenomenon of "rainbow stars", which you can see in the inset of the smaller star. This is an artifact that is very difficult to correct.

 

For the "final" M13 image I used the GHS stretch as a basis followed by stretches to bring a bit of contrast into the image, increased saturation on the globular cluster and a bit on the stars, desaturation on the background and local stretches to bring out the deteils in the core of M13.

 

 

First off, Fantastic Image!   I love images that I get lost in, and this is certainly one of them.   Astronomy question - Are those dust lanes in the foreground of the cluster are within it itself?   If within the cluster, than what is the difference between a cluster and mini/dwarf galaxy?   

Thanks also for the star analysis.   That helps me a lot!   

Mathematically, the GHS stretch is most like the masked stretch as well (and I believe is the actual limit if an infinite number of mask-steps were taken, and an exponential stretch were employed).    As for the HT,  I have to say it would be difficult to make a better all-round stretching transform with essentially one input control parameter, but can struggle with star bloat and bringing out dim features.     The ASH is the equivalent,  in terms of brightness, is the equivalent of performing the GHS stretch with a negative b-factor, so if you like the star shapes that it yields you can use either method.   The colour enhancing magic that ASH provides actually has little to do with "arcsinh" itself, but has more to do with the manner in which the stretch transform is applied.   @Mike1485 and I (well, mainly Mike) is thinking of a way to either put a similar formulation into GHS, or just the ability to stretch saturation / luminance as an alternative to R,G,B.   Currently, I think GHS is good a preserving saturation or separately "stretching" a colour, but lacks the enhancing ability of ASH  (I actually enjoy the colour enhancement in your image). 

That's all fine and well, but your posting makes it real, and much appreciated.   Those stars are a tricky business!  Thanks again for sharing!

[mike1485]

11 hours ago, Annehouw said:

To give a better idea of what GHS can accomplish, I made a compilation of 4 stretching methods:

1. Generalised Hyperbolic Stretch (GHS),
2. Masked Stretch (MS),
3. Histogram Transformation (HT)
4. Arcshinh stretch (ASH), which is actually a combination of a very mild HT stretch followed by ASH (otherwise it becomes a bit ugly...).

The stretches are not 100% equivalent, but it gives an idea.

 

The inserts show the bright yellow star on the right and a star to the left of this star. 

As can be seen, GHS looks a lot like MS. MS retains a bit more color, but the downside of MS is that it has created a harsh plateau in the star brightness, resulting in a distracting dark ring in the bright star. 

HT puts a lot of light into the halo and gives a relatively contrasty image (to be corrected by Local Histogram Transormation or HDRMT).

ASH is the colorful one. That is nice. Unfortunately, it also gives rise the the scary phenomenon of "rainbow stars", which you can see in the inset of the smaller star. This is an artifact that is very difficult to correct.

 

For the "final" M13 image I used the GHS stretch as a basis followed by stretches to bring a bit of contrast into the image, increased saturation on the globular cluster and a bit on the stars, desaturation on the background and local stretches to bring out the deteils in the core of M13.

 

 

That is a stunning image, and some excellent and very useful "real world" analysis/comparison.  Thank you so much for sharing this.

[Annehouw]

22 hours ago, Gunshy said:

Astronomy question - Are those dust lanes in the foreground of the cluster are within it itself?   If within the cluster, than what is the difference between a cluster and mini/dwarf galaxy?   

 

I was a bit lazy in my wording on "dust lanes".  Many of the dark structures seen in the image could just be differences in star density.

 

In general Globular Clusters are made up of a rather homogeneous collection of very old stars with not a lot of elements heavier than H and He ("metal-poor"). If there is dust and whether the dust is foreground or in the cluster itself is a bit tricky to answer. The best way to go about this would be to look at it in infrared wavelengths and look for dust re-emission of starlight. The dust in the disk of the galaxy is heated to about 20K due to the overall radiation. In Globular clusters, the stars are tightly packed and as a result of that there is more intense radiation. Dust (if any) in Globular Clusters should have a temperature of 50K-80K by that heating. So, the IR signatures of dust in Globular Clusters and that of dust in the Galactic Disk should be different. Whether this has been done for M13, I do not know. I did find that the Spitzer IR telescope has had a look at M15 and indeed found dust, but it is a very minute amount (something about  0.0001 solar mass for the whole of M15). The most likely source of that dust is the wind of stars in late evolution stages (Giants) where a lot of material of the outer shell is blown away, which will contain some dust.

As for dwarf galaxies, they have a more mixed polulation of stars and also a lot more gas (and a complete different evolutionary history.

 

Hope this is interesting and answers your question

 

Edited January 6, 2022 by Annehouw

[Gunshy]

11 hours ago, Annehouw said:

I was a bit lazy in my wording on "dust lanes".  Many of the dark structures seen in the image could just be differences in star density.

 

In general Globular Clusters are made up of a rather homogeneous collection of very old stars with not a lot of elements heavier than H and He ("metal-poor"). If there is dust and whether the dust is foreground or in the cluster itself is a bit tricky to answer. The best way to go about this would be to look at it in infrared wavelengths and look for dust re-emission of starlight. The dust in the disk of the galaxy is heated to about 20K due to the overall radiation. In Globular clusters, the stars are tightly packed and as a result of that there is more intense radiation. Dust (if any) in Globular Clusters should have a temperature of 50K-80K by that heating. So, the IR signatures of dust in Globular Clusters and that of dust in the Galactic Disk should be different. Whether this has been done for M13, I do not know. I did find that the Spitzer IR telescope has had a look at M15 and indeed found dust, but it is a very minute amount (something about  0.0001 solar mass for the whole of M15). The most likely source of that dust is the wind of stars in late evolution stages (Giants) where a lot of material of the outer shell is blown away, which will contain some dust.

As for dwarf galaxies, they have a more mixed polulation of stars and also a lot more gas (and a complete different evolutionary history.

 

Hope this is interesting and answers your question

 

Thanks so much for that great answer.   I know the question was a little off-forum topic, but I appreciate the answer.  I will definitely be looking for these "dust lanes" in my globular images.

[Lee_P]

Thanks for putting this all together @mike1485 and @Gunshy. I've just been trying it for the first time on some fresh M45 data. Here are some comparisons using three different stretching methods:

 

Data stretched by applying STF to the Histogram:

 

Data stretched using EZ Soft Stretch:

 

Data stretched using GHS:

 

GHS obviously has a lot of potential! There's loads more nebulosity, and the stars are nicely controlled too. And this was just my first play around with it. I'm processing the data now, but already I've discovered one happy result of using GHS: stars are so well controlled that Topaz Denoise doesn't have trouble with them in the same way it often does after stretching with the other methods.

If the developers are open to requests, I'd be very interested in a more user-friendly tutorial on the interface. I relied on the Zoom recording posted on YouTube, and that was just enough info to start me off. It would also be wonderful if there could be a real-time preview of your adjustments onto the actual image, rather than just a histrogram. No idea if that's possible.

Anyway, thanks again for making this tool!

 

Edited January 13, 2022 by Lee_P

[mike1485]

20 hours ago, Lee_P said:

Thanks for putting this all together @mike1485 and @Gunshy. I've just been trying it for the first time on some fresh M45 data. Here are some comparisons using three different stretching methods:

 

Data stretched by applying STF to the Histogram:

 

Data stretched using EZ Soft Stretch:

 

Data stretched using GHS:

 

GHS obviously has a lot of potential! There's loads more nebulosity, and the stars are nicely controlled too. And this was just my first play around with it. I'm processing the data now, but already I've discovered one happy result of using GHS: stars are so well controlled that Topaz Denoise doesn't have trouble with them in the same way it often does after stretching with the other methods.

If the developers are open to requests, I'd be very interested in a more user-friendly tutorial on the interface. I relied on the Zoom recording posted on YouTube, and that was just enough info to start me off. It would also be wonderful if there could be a real-time preview of your adjustments onto the actual image, rather than just a histrogram. No idea if that's possible.

Anyway, thanks again for making this tool!

 

Hi Lee

Thanks for sharing these images.  They give a really great comparison of various stretching methods.  I am glad you are finding the tool useful and really interested to hear your observation regarding Topaz Denoise.  It's not software I use but I know a lot of people do!

Thanks also for your suggestions for future development.  We are working on a "version 2" so it is always helpful to know what people would find useful (without any promises I'm afraid - its not our day jobs!)

All the best and clear skies!

Mike

[Lee_P]

6 minutes ago, mike1485 said:

Hi Lee

Thanks for sharing these images.  They give a really great comparison of various stretching methods.  I am glad you are finding the tool useful and really interested to hear your observation regarding Topaz Denoise.  It's not software I use but I know a lot of people do!

Thanks also for your suggestions for future development.  We are working on a "version 2" so it is always helpful to know what people would find useful (without any promises I'm afraid - its not our day jobs!)

All the best and clear skies!

Mike

Let me know if you want a Beta tester for version 2 😃

Thanks again for making and sharing GHS!

[Gunshy]

I just wanted to show a revision I made to some year old NB data I had on NGC2337 (Rosette).  I was able to really bring out the nebulosity around the Rosette and control stars using GHS.  To avoid bringing out noise as well, Mure noise reduction helped tremendously. 

(The handling of the stars was my own experimentation with Starnet to substitute RGB and I used the method of Cuiv the Lazy Geek to achieve my take on the Hubble pallette)

[mike1485]

On 26/01/2022 at 18:12, Gunshy said:

I just wanted to show a revision I made to some year old NB data I had on NGC2337 (Rosette).  I was able to really bring out the nebulosity around the Rosette and control stars using GHS.  To avoid bringing out noise as well, Mure noise reduction helped tremendously. 

(The handling of the stars was my own experimentation with Starnet to substitute RGB and I used the method of Cuiv the Lazy Geek to achieve my take on the Hubble pallette)

Hi Dave - this is just stunning  !

[mike1485]

I wanted to make all those watching this thread aware that I have today released version 2 of GHS.  Amongst the improvements included in this update is a real-time preview functionality.  I have started a new thread for the new release here: GHS Version 2 thread.

CS, Mike