Page 1 of 1

masked stretch

Posted: Thu Feb 07, 2019 8:39 pm
by Rkonrad
I find it challenging to stretch star rich data as it blows up the roi plus stars. I have devised a way partly around this by applying an inverse star mask, then applying "isolate". It's a bit of a crude technique and takes a lot of time and experimentation. In the end it kind of works if the mask is perfect yet always applies a blur whether I want it or not (airy disk radius set to maximum). The moderator (Blair McDonald) of the dslr yahoo group uses Images Plus and uses a star mask to stretch data from the start. Here's an example of his go at the current challenge image. https://www.flickr.com/gp/rkonrad/Fs00um Data is from a dslr no HA.

Thanks for your help!

Richard

Re: masked stretch

Posted: Thu Feb 21, 2019 12:54 am
by admin
Hi,

Could you elaborate on the technique, the result you're after and the result you're getting instead?
What do you mean by 'blows up the roi plus stars'?

Currently, the image you have linked to does not exists...

Thanks!

Re: masked stretch

Posted: Thu Feb 21, 2019 3:58 pm
by Rkonrad
Hi Ivo,

When stretching a star rich image, the stars increase in luminance as well as the main dso. Star shrinking in the end only works so well unless the stars are reasonably in control from the start I find. Here is Blair McDonald's version of the Yahoo group challenge. https://www.flickr.com/gp/rkonrad/Y4Q6VT. Notice the how small the stars are. He uses Images Plus.

This is my version https://www.flickr.com/gp/rkonrad/45V206 which I achieved by doing a inverse star mask, then isolate with the air disk radius set at max. This separated the stars from the dso by increasing luminance of the nebula and reducing the stars slightly. It works but its also a lot of work as the stars have to be masked absolutely perfectly for the them to look natural. It also gives a slight blur whether I want it or not.

Hope that's enough info for you. Cheers Richard

Re: masked stretch

Posted: Fri Feb 22, 2019 1:44 am
by admin
Rkonrad wrote:Hi Ivo,

When stretching a star rich image, the stars increase in luminance as well as the main dso. Star shrinking in the end only works so well unless the stars are reasonably in control from the start I find. Here is Blair McDonald's version of the Yahoo group challenge. https://www.flickr.com/gp/rkonrad/Y4Q6VT. Notice the how small the stars are. He uses Images Plus.

This is my version https://www.flickr.com/gp/rkonrad/45V206 which I achieved by doing a inverse star mask, then isolate with the air disk radius set at max. This separated the stars from the dso by increasing luminance of the nebula and reducing the stars slightly. It works but its also a lot of work as the stars have to be masked absolutely perfectly for the them to look natural. It also gives a slight blur whether I want it or not.

Hope that's enough info for you. Cheers Richard
Hmmm... it appears Flickr's image storage is down at the time of this writing (no images showing up, getting 502 Bad Gateway server errors).

Instead I went into the Yahoo Group to look for Blair's image and more information on his work flow. I'm still a bit hazy on what you mean by "masked stretching" or how it would be applied to this image? What would masked stretching do in your own words? And why would it be useful in your case?

Correct me if I'm wrong, but Blair's workflow seems to just separate stars entirely and just blends them in at a later stage? From the Yahoo Group;

1. Arcsinh stretch
2. Separated stars and nebula
3. Made a starless mask from the nebula layer
4. Restretched the original image using the mask
5. Fixed the stacking artifacts with a lighten combine to keep as much of the image as possible
6. Masked noise reduction
7. General stretch functions n’th root mode
8. Feature mask to split the stars and nebula
9. Tweaked the colour of the nebula
10. Sharpened the nebula layer using multi-resolution sharpen
11. Some manual sharpening using the sharpen brush
12. Upped the saturation of the star layer
13. Recombined the nebula with the stars
14. One pass of star shrink
15. Cropped
16. Binned by two
17. Set black point with levels

This technique can just as well be accomplished with StarTools (using Heal module and Layer module); create Star Mask, Heal, Launch Layer module, Undo->Foreground, select Brightness Mask Mode 'Where Composite is dark, use background', then play with the Brightness Mask Power.

You can also use the Life module's Isolate preset in a subtly different way. As you know it's great for pushing back busy star fields. So let's enhance that behaviour! :twisted:
Try the following (after Color calibration, so you can better see how the following technique will be affecting star halos)
Life Isolate preset without mask, Keep
Layer, Undo > Bg,
Mask -> Auto, Fat Stars, Do, Grow x2 or x3 (to cover the halos), Keep
Back in Layer module you will now get the "pushed back stars" with the original nebulosity.

You can "overdrive" the effect as well; increase "Blend amount" beyond 100% (gaps will form) and enhance increase "Mask Fuzz" to make the blend smooth again.
You can further experiment with Brightness-dependent masking, for example setting Brightness Mask Mode to "Where composite is light, use bg" and playing with the Brightness Mask Power in conjunction with the Mask Fuzz. Finally, you can repeat this whole process after 'Keeping' the result from the Layer module doing another iteration with the Life Isolate preset, etc.

It's interesting not many people seem to care about the coloring on the DSLR group, which is a shame. It is clear this dataset has been shot with some sort of filter, so any attempts at recovering "visually accurate" coloring wouldn't be very fruitful. However, it is really easy to show the different types of light/emission (and variations in star temperatures) that were recorded and made it through the filter. The area is strong in Ha emissions, but also OIII emissions. (see also http://www.astronomersdoitinthedark.com ... =151&p=520). Trying to shoehorn everything into the red channel in the futile pursuit of trying to establish "visually correct coloring", even though OIII firmly belongs in the green channel is a shame IMHO. OIII emissions (green) are readily visible in this dataset - StarTools' default Color calibration teases its out nicely;
C060_HeartNebula.jpg
C060_HeartNebula.jpg (250.36 KiB) Viewed 8224 times
Hope this helps!

Re: masked stretch

Posted: Fri Feb 22, 2019 1:44 pm
by happy-kat
Interesting thread, some ideas there. Thanks.

Re: masked stretch

Posted: Sat Feb 23, 2019 5:34 pm
by Guy
Yes, very interesting.
The challenges from the DSLR-AIP Yahoo Group, like this one with data provided by Marco Labor, are a great way to learn new techniques.

Ivo, do you have the StarTools session log for that approach? I'd like to try and recreate it using the latest alpha of STReplay.

Guy

Re: masked stretch

Posted: Mon Feb 25, 2019 7:52 pm
by Rkonrad
Thanks Ivo!

This is an important issue for me but I don't have time to respond properly just now. I'll be in touch.

I'm giving a presentation on Startools at my local astronomy club (RASC) early in March. I respond after that - though I'm likely to have a few more questions of a more general nature about the programme.

Cheers

Richard

Re: masked stretch

Posted: Tue Feb 26, 2019 1:10 am
by admin
Guy wrote:Yes, very interesting.
The challenges from the DSLR-AIP Yahoo Group, like this one with data provided by Marco Labor, are a great way to learn new techniques.
Definitely, datasets like those of Marco are brilliant examples of "real life" datasets with some unique challenges.
Ivo, do you have the StarTools session log for that approach? I'd like to try and recreate it using the latest alpha of STReplay.
I don't think the Layer module outputs all button presses to the log yet... :(

This is, using the Kurt M. dataset, a demonstration of the first technique (e.g. create star mask, heal, layer back in undo buffer with brightness masking)
scrshot.jpg
scrshot.jpg (294.89 KiB) Viewed 8159 times
The second technique (using one or multiple iterations of the Life module);

--- Life
Clear mask.
Isolate preset.
--- Layer
Undo -> Background
Parameter [Brightness Mask Mode] set to [Where composite is light, use bg]
Parameter [Brightness Mask Power] set to [2.20]
Parameter [Blend Amount] set to [273 %]
Parameter [Mask Fuzz] set to [15.0 pixels]
Rkonrad wrote: This is an important issue for me but I don't have time to respond properly just now. I'll be in touch.
I'm giving a presentation on Startools at my local astronomy club (RASC) early in March. I respond after that - though I'm likely to have a few more questions of a more general nature about the programme.
No hurry - best of luck with the Preso!

Re: masked stretch

Posted: Wed Feb 27, 2019 10:31 pm
by Guy
Hi Ivo,
Thanks for the clarification.
I know its difficult to represent some of the actions in the log file - I wanted to try some STReplay commands which may support these more complex tasks.
I'll let you know how I get on.