Messier 106 et al.

[Mike in Rancho]

Maybe I will drop a post in Mike's thread. He's so silent

Not sure if serious.

I have been relatively quiet lately. Work, errands, commuting, chores, clouds, new cat condo, and maybe a mild case of Corned Beef and Cabbage Coma.

Enough housekeeping.

A number of interesting points are raised in this thread. Not sure it's derailing the M106? Many modules have been discussed, and I know you are also interested in looking under the hood to find out what tools are actually doing. I'm glad some of them have been brought up, since I have been pondering them as well. Andy raising AutoDev/OptiDev matters dovetails nicely too, so I'll divert to that momentarily...


@AndyBooth ,

Nice images and interesting workflow explanation. Apart from you, I'm only aware of a handful of folks who have come up with a way to blend ST with other software, successfully. So that's cool. Of course I mean substantively too, not just some tweaking after the fact in PS, Gimp, or whatever.

Some of that may be because ST does not play well with others. Designed around tracking is but one aspect. The other is probably the 16-bit TIFF RGB as the only real (that I know of) save-out option. I wonder if combo usage involving ST would have broader appeal if we had more and higher depth save options. There are many tools in ST that are either unique, a couple of which you mentioned, or more advanced, such that people might find them very useful a-la-carte, even without tracking.

The stretch being one of your main points of diversion is intriguing as mentioned, and I'll jump back to Dietmar for that.

However, I don't know if you are necessarily throwing out documentary or "scientific" value just by missing out on tracking. Of course ST is built for that to help with NR and for the modules to cross-talk via time-shifting. But other software can be used in a straight up manner as well. A different global stretch to kick things off seems reasonable to me.

On star removal, I suppose the devil is in the details of how it is applied. Your write-up appears to signify use as more similar to perhaps shrink, or piecing together aspects that are both real, but in need of different non-linear stretching. Worthy of some pondering.


@decay

Some of this could be moved to the AP/Art/Experiment thread for deeper discussion as you say, though this thread has raised the issues also.

I have wondered myself if the sometimes "harsh" appearance of data in ST is more than just the noise view applied at sub 100% scale? Many times even autostretches, whether in PI, Siril, or apparently APP, without even getting to GHS, just look "better," and again query how much of that might be downsampling of the view (binned SNR).

I have been considering if range setting, whether in Wipe or OptiDev, meaning black point mostly, could be at the root of things. Stretches in other software often leave a bit of a pedestal on the left side of the histogram. I believe that may "soften" the appearance, and also may hide noise within the brighter overall background. FilmDev offers a possible pedestal (skyglow), but not OptiDev unless there's another way to achieve that.

I am thinking of making some sample images for testing stretches, similar to the ones you made for Sharp experiments. ST's dynamic stretching may make it a bit difficult, but perhaps still useful with some playing around. Maybe just a grayscale set of vertical bars at intervals, like from 0 to 255. I saw something similar in an Adam Block video on GHS, and it was quite interesting.

That could possibly be used in conjunction with the Dietmar app? Otherwise, as I think you alluded to, I think it's difficult to get a graphic visualization of what changed if there's only a before and after state, and no live histogram within the program.

And I still want to dig into what OptiDev is doing with saturated star cores, and why the shape of the PSF ends up so different from FilmDev or other software stretching.


I'll see what I can make up when I get some time. Back to being silent....

PS - Oh, I have to get myself back on A-bin so I can hit the Like for you. My subscription expired and I haven't had anything worthy of uploading to bother renewing yet.

[AndyBooth]

@Mike in Rancho

On star removal, I suppose the devil is in the details of how it is applied. Your write-up appears to signify use as more similar to perhaps shrink, or piecing together aspects that are both real, but in need of different non-linear stretching. Worthy of some pondering.

Yes , exactly!
The processing of the object well, always over processes the stars in my view, although sometimes unavoidable, I am not a fan of a harsh defined core and then halo, but at least try for a softer graduated star, but still retaining colour (wherever possible).
So I do the starless thing mainly to replace the stars with the same stars from the same data but only very mildly stretched, and well saturated.
However, on duo band / bi colour renditions, I also do starless to replace the stars with ones taken from broadband data, to make them look natural, again it prefer it to the just white, or just bi colour stars.

and I know you are also interested in looking under the hood to find out what tools are actually doing.

Yes, I am too!. I still say that every deepsky image that I have processed since its inception has used startools modules, and still find it an astonishing piece of software for bringing out detail etc (synthetic lum processing etc). In my heart I would love to help contribute to a process that tackles some of the issues (just personal likes and dislikes- not faults) I use other software to tackle.

To this end, I intend to do parallel processing over my next images, with better documentation, 1. With fully startools compliant APP settings and full startools processing, and 2. My normal mash up. See I can offer any thing up.


Best regards,
Andy

[AndyBooth]

Regarding stretches,
Here is the ‘innovative’ method now used in both Pixinsight and Siril,
The General Hyperbolic Stretch method.
As Startools in Optidev tries to balance darks, mids, and highs with no overall bias to either, I wonder what stretch it uses,
Of course it selects its own optimum parameters automatically I guess.

https://siril.org/tutorials/ghs/

[decay]

Thank you all! Much appreciated. My intention when posting images is not just to do that, but perhaps to start some conversation as well. I guess I succeeded in this case. Talking about what we are doing is important in my eyes and why not start it by posting an image?

So no one has to be afraid of derailing this thread. It’s part of my intention. But right now it’s too much at once. I will reply to all your posts in the next few days. Hopefully.

@Mike in Rancho : I would be glad to continue to look under the hood. Of course, always as time allows for both of us (or anyone else who wants to participate of course). And we should do that step by step and starting with OptiDev / global stretches is just fine for me. We could start a new thread to keep things separated. What do you think?
And please don’t do anything right now. (Please do not draw 256 bars using the Gimp! ) I already have some ideas and will write them down soon.

Best regards, Dietmar.

[Mike in Rancho]

Yes , exactly!
The processing of the object well, always over processes the stars in my view, although sometimes unavoidable, I am not a fan of a harsh defined core and then halo, but at least try for a softer graduated star, but still retaining colour (wherever possible).

....

To this end, I intend to do parallel processing over my next images, with better documentation, 1. With fully startools compliant APP settings and full startools processing, and 2. My normal mash up. See I can offer any thing up.

Regarding stretches,
Here is the ‘innovative’ method now used in both Pixinsight and Siril,
The General Hyperbolic Stretch method.
As Startools in Optidev tries to balance darks, mids, and highs with no overall bias to either, I wonder what stretch it uses,
Of course it selects its own optimum parameters automatically I guess.

It can do so, yes. One of the reasons I'd like to dig into what OptiDev is doing up in the high range. I actually appreciate a pinpointed star core with a diffraction ball. Perhaps moreso with my Newt. I can understand it maybe looking more "unnatural" with refractor stars. But there can be issues, such as a flat white disk following SVD as Martin has alluded to a few times. I want to experiment to see if this is borne out of earlier OptiDev stretching.

But even though I like the tighter cores, I also still want there to be a graduated PSF, rather than a volcano or donut which seems to sometimes happen.

Your full workflow comparisons will be interesting!

Oh and yes I have played some with GHS in PI, and maybe once in Siril. While I think an improvement, it is still tedious iterative histogram work. But OptiDev manages to bring out detail in one go for a single global initial stretch curve, that would take multiple iterations in GHS to tease out. But as you note something is off sometimes, possibly a black point issue.

I don't know if, aside from the custom controls of course, it is true that ST is not biasing amongst shadows/mids/highlights. It is detail dynamic, but I am insure what it considers the "optimal" curve.

And we should do that step by step and starting with OptiDev / global stretches is just fine for me. We could start a new thread to keep things separated. What do you think?
And please don’t do anything right now. (Please do not draw 256 bars using the Gimp! ) I already have some ideas and will write them down soon.

I wasn't going to use a step size of 1. That would just be...a gradient?

But yes several test panels, including some that start and end short of the extremes, to see what OptiDev (or Wipe if we need to try to squeeze that in) is doing with the overall range.

Whether it goes into the same experiment thread or not doesn't matter. I think it's reasonable to include and at the beginning I also did mention where various stretching techniques might fall.

Though that reminds me too - do you remember if there was a "in the works" features and docs page for 1.9? I thought there was but maybe I dreamed it. I'm still using the main features and docs for my background reading.

[decay]

Hi Martin @Startrek ,

thanks for your reply . I cannot quote, because this thread is too long and your text is already out of reach. So I will quote the old school manual way

Martin wrote: "I’ve persevered with 1.8 SV Decon for over a year and have reluctantly switched back to 1.7
I just can’t avoid having my stars cores ( big and small ) ending up looking like super white discs. Whilst the Deconvolution in SV Decon works extremely well ( deblur ) in comparison to 1.8 synthetic or 1.7 PSF , the stars end up looking artificial. Maybe I should try 1.9 and see if I can make the stars cores look more natural. In your image the larger star cores show this white disc but the smaller ones look more natural."

Yes, I've followed your journey back to 1.7. Maybe we can elaborate some things when analysing OptiDev. I'm sure there are some changes compared to prior versions. If I remember correctly, Steve had some trouble with stars showing dark(er) cores after HDR. I did a comparison to 1.8 and indeed this effect was much weaker. And maybe there are consequences for SV Decon as well. Who knows? Going back to 1.7 is too bad - it's frustrating for you, but I'm sure for Ivo as well. A lot of work and dedication went into 1.8 and 1.9 and I'm sure it is difficult to consider all aspects and expectations. Maybe we can help with analysing and testing. At least a bit. Too bad that Ivo is busy for such a long time.

Martin wrote: "By the way I’ll swap your B5 skies for my B8 skies any day , B8 is hard work and broadband with my 2600MM is virtually impossible unless you want to process hundreds and hundreds of 30 sec subs , I don’t have consistent clear skies or storage space to efficiently capture galaxies under my B8 , I’ll leave that to my rural site where my Dome is located under B3."

Of course, I should not complain. The most important thing is that I can do this at home and don't have to drive to a dark place. I would never do that, because I don't have enough free time. And of course I'm too lazy

But maybe one fine day I will have my own Dome located under B3, too. At least I can dream about it

Best regards, Dietmar.

[AndyBooth]

@Mike in Rancho
Yes, I agree GHS is very tedious to use indeed, and yes, its really hard to get all stars to look natural whichever software and method you use.
I have had more success with the starless approach, but of course the tight stars from svdecon are so beautifully small too!
Aghh.
You are right, stars from a newt or RC are what I grew up on, i’ve had several newts and an RC, and I only got a frac beginning last year, and its a learning curve on stars.
But I had a difficult time with newts even though I’ve ground mirrors and used them in the 1970’s as a teenager.
I attach a newt pic from a couple of years ago, whatever I tried, I could not get rid of the poor diffraction pattern from spider and clips etc, even though using a coma corrector. This was almost all startools processing, including initial stretch, with only minor tweaks in PS. Not starless process or anything.
I am enthused to really give Startools a real crack now with the frac, I feel I am supported by this forum and its very helpful input.
Thx for your input.

Ps. still not got the posting rules sorted for best quality lol.

Best regards,
Andy

[decay]

Though that reminds me too - do you remember if there was a "in the works" features and docs page for 1.9? I thought there was but maybe I dreamed it. I'm still using the main features and docs for my background reading.

Sorry, Mike. I can't remember and a quick search and scanning of the 1.9 preview thread didn't turn up anything.

I will reply to our other points and topics later.

Dietmar.