StarTools

Hi Ivo (and all of course),

there was one sentence in your post
viewtopic.php?p=13385#p13385

that caught my special attention:
“… and the signal has to be pushed quite a bit. The latter can make noise start to poke through ...”

Together with this post (explaining how the NBAccent module works)
viewtopic.php?p=12382#p12382

and particularly this point
“the new value for the pixel (per channel), is the value of the pixel that is largest (original or accented)”

I’ve started thinking about what this means in case of noisy data sets.

If I understand correctly this means for every single pixel
- we have two signals (original or accented) with different levels
- both signal levels are superposed with different, random noise levels (shot noise or whatever)
- therefore the outcome (which signal “wins”) relies on the (for each pixel) concrete, random, statistical noise levels both signal levels are superposed with.

This means if there is for example an area in original data with signal level(s) not too much above the signal level of the accented data and both signals are noisy that there are a number of pixels in this area where the accented data “wins” – due to random noise levels. So the accented data shines or pokes through – as you wrote.

Do I understand this correctly? (I hope what I wrote was halfway understandable. Not so easy to describe.)

Having all this said, here comes my question: Would it make sense to do first noise reduction on both data sets (separately) and than afterwards NB accent processing as the ST module does? Having lower noise levels on both data sets should reduce this ‘poking through’ effect?

Thanks & best regards, Dietmar.

decay wrote: ↑Fri Oct 21, 2022 6:06 am I’ve started thinking about what this means in case of noisy data sets.

If I understand correctly this means for every single pixel
- we have two signals (original or accented) with different levels

Correct.

- both signal levels are superposed with different, random noise levels (shot noise or whatever)

Correct.

- therefore the outcome (which signal “wins”) relies on the (for each pixel) concrete, random, statistical noise levels both signal levels are superposed with.

Correct.

This means if there is for example an area in original data with signal level(s) not too much above the signal level of the accented data and both signals are noisy that there are a number of pixels in this area where the accented data “wins” – due to random noise levels. So the accented data shines or pokes through – as you wrote.

Do I understand this correctly? (I hope what I wrote was halfway understandable. Not so easy to describe.)

Correct. Of course, the signal is first transformed (stretched, thresholded) and modulated in various other ways by the non-accented signal.

Having all this said, here comes my question: Would it make sense to do first noise reduction on both data sets (separately) and than afterwards NB accent processing as the ST module does? Having lower noise levels on both data sets should reduce this ‘poking through’ effect?

At the very start of NBAccent's research and development, that's what I thought as well.

It turns out though, that Tracking does a much better job on the combined signal during final noise reduction, rather than on just the NB signal before compositing in the NBAccent module. It makes sense when you think about it; the signal flow is setup such that Tracking regards the NB signal as "not there" in the final image, meaning that any shot noise (e.g. non-correlated noise) that is introduced by the accents is completely obliterated in the final noise reduction stage, if it thinks the non-accented signal/detail suffers. The accents are truly in the service of the original detail from a signal processing point of view.

This is also the reason why using the NBAccent module should yield much better final results than any traditional compositing technique, and why you should be able to use very marginal data to enhance your visual spectrum image with accents.

The latter is very often the case when dealing with Ha accents of galaxies (as I mentioned in that post); there is usually very little signal except in a few "knots".

For a good example of such marginal data see the Elf's Cloudy Nights data. One of my most challenging test datasets was his Pinwheel galaxy. Visual spectrum (intentionally left as luminance only) looks like this; But the Ha data looks like this; The background is very uneven and thresholding is needed. All said and done though, after compositing noise is very much under control; and final noise reduction removes any remnant noise very effectively as described above;
You can clearly see that if there is no detail in the original signal in a location, then noise reduction all but obliterates the noise blotches etc. from the NB accent signal.

Hope that makes sense and helps!

That's cool! And a great explanation.

For all the Elf data I did from that thread, I believe I skipped M101. But, I don't think NB Accent existed then either.

Hi Ivo,

thank you for your detailed answer!

admin wrote: ↑Sun Oct 23, 2022 12:58 am the signal flow is setup such that Tracking regards the NB signal as "not there" in the final image, meaning that any shot noise (e.g. non-correlated noise) that is introduced by the accents is completely obliterated in the final noise reduction stage, if it thinks the non-accented signal/detail suffers. The accents are truly in the service of the original detail from a signal processing point of view.

That's fantastic. So again, it's an achievement of signal evolution tracking.

There have been two major reasons for me to choose StarTools. One was the WIPE modul. Other tools drove me crazy and using them was always kind of unsatisfying. But what is really outstanding in my eyes is signal tracking. When you think about it it's so evident: Of course in image processing workflows every pixel has a history. And this history has to be considered in further processing. And further on modeling the underlying physics of the signal and the impact of processing steps on this signal is obviously what we want do to. And not just unrelated pixel maths.

Simply excellent. Nothing more to say on top.

Best regards, Dietmar.

Of course I've more to say

admin wrote: ↑Sun Oct 23, 2022 12:58 am For a good example of such marginal data see the Elf's Cloudy Nights data. One of my most challenging test datasets was his Pinwheel galaxy.

If you are ever looking for limited data sets, don't hesitate to contact me! I'm able to provide data sets with all kind of imperfections

Here are some screenshots of my M31 processing. First one is luminance only: NBAccent with default settings. There is a lot of red hue everywhere - only the core remains white: NBAccent with adjusted parameters. For me probably the best tradeoff I got so far:

And last the the one with adjusted parameters after NR: Just preliminary here to give an impression what was bothering me. I will post a more detailed description (and data sets) under 'Troubleshooting' in the next few days, maybe someone can help ... or tell me, that my data is just too bad

Best regards, Dietmar.

Hi All,

Most interesting thread is this.

Me too struggling. My Ha is also very noise, obviously cause i use thesame technique as the Elf for taking Ha.
the threshold parameter in NBaccent seems to do indeed a tremenduous job to get the noise out but now i can't get any color in my Ha regions.
You can see them but they remain white.
Object is M81.

Hi Freddy,

my specific case was discussed a few days later is this thread:
viewtopic.php?f=10&t=2721

Ivo explained that some processing can easily be disturbed by correlated noise and therefore NBAccent did not work well for me in this case. Using the correlated noise filter in Wipe helped indeed quite a bit.

Hope this helps.

Best regards, Dietmar.

Thanks Dietmar,

I know about the thread, have to read it still through though.
Meanwhile i was able to get some pink/red in the ha regions by using another setting in the compose module. But as your M31 i winded up with pink/red in the center of M81 too. I think that might be expected since in the Ha stack the center is recorded too.

I think i am on the right track but will need quite some excercie before i get something descent. First time i try this. It was a bit of a challenge to frame the two galaxies too, since with the Ha filter in place much light is blocked...

Hey Freddy, what are you using for acquisition? Off the top of my head it's APT or something like that? No plate-solving?

It may take some research and experimentation, but NINA should work with most DSLR - at least Nikon and Canon. I don't think it's quite as clean with Nikon as say, BYN, but the added features are nice including the framing wizard.

What is the Elf's technique for taking Ha?

Getting Screen 1 of NB Accent is of course key, and I often hit the back button to re-do it. Everything depends (I think) on pixel size of Ha regions passed over, and thus sometimes the Galaxy preset doesn't show off all of the transferred Ha. Adjusting the two bottom sliders on the right - brightness correlation and detail size (?) can help with that, if Screen 2 isn't showing much if any of the Ha. As can strength on the left side.

For me though I haven't got it to a point (if possible) where I don't then do a Layer/Undo reversal on continuum light from the stars, occasionally overall background and/or inner galaxy disks, to remove a pink donut hole.