Combining OSC with Red only (Ha) from L-extreme

[admin]

Hi Ivo,

I'm trying to capture the technique you describe here and I have a few questions.

- If I just load an OSC/DSLR file into the luminance channel (and nothing into R,G or B) and set 'Luminance, Color' to 'L + Synthetic L from R(2xG)B, Mono (Mono from OSC/DSLR)' does it extract the RGB from the OSC file to create the synthetic luminance data?

Whenever you import a file that contains multiple channels, it will extract any relevant channel (for example, it will extract the red channel if you are loading it as "red", etc.). In the case of importing an RGB file for your luminance, it will simply sum (average) the three channels (e.g. it will not do any re-weighting).

- I do this with the wide band OSC file and with the filtered Ha data file (presumably also from the OSC) and create luminance data files.
- If I use the Compose module to combine the two files - say I use R channel for Ha & G channel for Wideband and I set set 'Luminance, Color' to 'L + Synthetic L From RGB, Mono':

• Do I set Channel Interpolation to Off since we don't want a synthesised B channel adding to our synthetic luminance ?
• Do I set 'Red total exposure' to 312 and 'Green total exposure' to 228 and 'Blue total exposure' to 'Not Set' to get the weighting of the two sets of luminance data correct?

Channel interpolation purely impacts coloring only. It does not add/impact/modify synthetic luminance.

So I then have the luminance file and load it into Luminance channel and load the wideband OSC file into the R G and B channels individually.

• Do I set 'Luminance, Color' to 'L,RGB' and leave exposure times to 'Not set' as R,G and B all have same exposure time?

Indeed. The exposure times don't do anything really when L, RGB is selected (since there is no synthetic luminance being generated). I need to disable them in that case! (it's on my list )

[*] Is there any sense in setting 'Luminance, Color' to 'L+ Synthetic L from RGB,RGB' since we have already extracted the synthetic luminance from the RGB?
[/list]

If exposure times are all "not set", then no (0%) synthetic luminance would be added to whatever was loaded as "proper" luminance.

One final unrelated Compose question:
- If I set total exposure time in a channel to zero it ignores the channel when creating synthetic luminance data - useful if you want to put the same (narrowband) data into 2 channels (e.g. OIII/Hb in G & B)- is that right?

Correct! This indeed allows you to decouple coloring from luminance and still achieve whichever luminance weighting you'd like.

Hope this helps!

[Guy]

Thanks Ivo - a great help.

Just to get it totally clear in my head...

Whenever you import a file that contains multiple channels, it will extract any relevant channel (for example, it will extract the red channel if you are loading it as "red", etc.). In the case of importing an RGB file for your luminance, it will simply sum (average) the three channels (e.g. it will not do any re-weighting).

So if I just load an OSC/DSLR file into the luminance channel (and nothing into R,G or B) and set 'Luminance, Color' to 'L + Synthetic L from R(2xG)B, Mono (Mono from OSC/DSLR)' there is no weighting R(2xG)B.
So do I need to load the OSC file into R, G and B as well in order to get the Synthetic L generated according to the weighting R(2xG)B?

So in the approach you describe - in order to create the required wideband luminance data file - rather than just load the OSC file into luminance only I should also load it into R,G and B to get the synthetic luminance generated from th setting of 'Luminance, Color' to 'L + Synthetic L from R(2xG)B, Mono (Mono from OSC/DSLR)' - is that right?

However, when creating the associated narrowband (Ha) luminance data file you can just load the OSC file into luminance as there is no significant G or B to create synthetic luminance from - is that right?

Finally...
Do I set 'Red total exposure' to 312 and 'Green total exposure' to 228 and 'Blue total exposure' to 'Not Set' to get the right weighting of the two sets of luminance data when combining the two luminance files?

Thanks,
Guy

[admin]

So if I just load an OSC/DSLR file into the luminance channel (and nothing into R,G or B) and set 'Luminance, Color' to 'L + Synthetic L from R(2xG)B, Mono (Mono from OSC/DSLR)' there is no weighting R(2xG)B.

Correct.

So do I need to load the OSC file into R, G and B as well in order to get the Synthetic L generated according to the weighting R(2xG)B?

The emulate precisely what happens when you use the second option (e.g. "Linear, from OSC/DSRL with Bayer matrix and not white balance"), you simply load the same dataset in R, G and B (which will extract the 3 channels and will weight green accordingly). There is no need/use for L. The entire luminance signal is/will be synthetic.

So in the approach you describe - in order to create the required wideband luminance data file - rather than just load the OSC file into luminance only I should also load it into R,G and B to get the synthetic luminance generated from th setting of 'Luminance, Color' to 'L + Synthetic L from R(2xG)B, Mono (Mono from OSC/DSLR)' - is that right?

You would not load the OSC/DSLR file into luminance (as it does not perform the correct weighting), but rather rely on the synthetic luminance generation. E.g. you would use 'L + Synthetic L from R(2xG)B, Mono (Mono from OSC/DSLR)' to create that synthetic luminance. Once you have that you can save it away. You can now choose using the Compose module or Layer module to make a blend of the Ha and synthetic L.

However, when creating the associated narrowband (Ha) luminance data file you can just load the OSC file into luminance as there is no significant G or B to create synthetic luminance from - is that right?

There are a few ways of adding the Ha (as luminance) to the synthetic OSC luminance.

Indeed, you could load the synthetic OSC luminance as L now, and load Ha into R, G or B. You could then use the synthetic luminance generation once more, in order to add the Ha you loaded into R, G or B as synthetic luminance. This will then be added to the synthetic OSC luminance you loaded into L. You would use 'L + Synthetic L from RGB, Mono (Mono from OSC/DSLR)' for this.

You could also fire up the Layer module, load the synthetic OSC luminance in the foreground (or background) and load the Ha into the background (or foreground). Set the 'Blend Amount' to whatever blend you need.

Finally...
Do I set 'Red total exposure' to 312 and 'Green total exposure' to 228 and 'Blue total exposure' to 'Not Set' to get the right weighting of the two sets of luminance data when combining the two luminance files?

Given that the Ha only lives in the (far) red channel, you would only set the red total exposure when adding its signal to luminance, and leave the other channels 'Not set'.

However, as stated in my post, there is no exact way to determine how much Ha should be added to the synthetic luminance; what might be appropriate/useful wholly depends on the object and your goals. SNR will likely increase for areas rich in Ha (good thing) but anything not Ha will be attenuated in terms of visibility (bad thing depending on the object) and such areas may/will receive some extra noise (bad thing depending on how much noise).

Ha tends to suffer less light pollution and skyglow noise (largely taking that portion of shot noise out of the equation), but other types of noise (read noise, thermal noise, etc.) will still be present and will get added.

It's a complex story with no clear-cut answer. The best advice I can give, is to have a clear goal in mind before acquiring narrowband signal to augment visual spectrum signal.

[Guy]

Thank you very much, Ivo, for your comprehensive reply.

I had lost sight of that fact that we were not talking about an approach that would produce documentary results.
I also forgot the luminance would be entirely synthetic.

I'll add the technique to the others in the Compose Module notes.

If anyone sees any mistakes please let me know.

Guy

[Mike in Rancho]

Oh I have lots of follow up and ever more questions.

So, based on the discussion, does this mean that the Ha + R technique in the ST manual and notes should not include putting the OSC file into L for that final compose? Seems so? And in order to get what I think the intended effect of that instruction was, requires more of a two-step process.

Unless it's just me (might be!), all or most of the special compose techniques could probably stand to mention settings for both the L,RGB setting and the channel interpolation setting. Unless silence can be understood to mean leave at default?

Channel interpolation, frankly, seems a bit mysterious, unexplained, and perhaps confusing by its labeling and description. Further detail would be handy. The stated function is to interpolate missing channels. But clearly there is more to this setting since one needs to have it "on" for mono channel extraction. That seems counterintuitive. Why would a mono red need missing green and blue to be interpolated? I have found that turning it "off" for single channel extraction will show it as the selected color (if a bit metallic looking).

My final "wish list" for compose info is a better understanding of what Synth L generation and L + Synth L means. Perhaps much is my minimal understanding of photo pixel array configuration and image processing "math" functions. Above it was noted as "sum (average)." ?? I presume it means that all signal, or perhaps signal strength, is preserved into the mono L regardless of whether it came from R, B, or G.

But if something more selective was thrown into L (such as Ha, commonly), couldn't that throw the L + Synth L combo a bit out of whack, or is that somehow handled by the blending?

Hope I haven't confused everyone as much as I have myself!

[Guy]

I'll answer the first question with regards to the Compose Module Use notes is concerned.

So, based on the discussion, does this mean that the Ha + R technique in the ST manual and notes should not include putting the OSC file into L for that final compose? Seems so? And in order to get what I think the intended effect of that instruction was, requires more of a two-step process.

The notes for the technique 'Adding Ha to OSC data' were derived from this thread Adding Ha data to OSC data which I have mis-transcribed. In the final step setting Set 'Luminance, Color' should be to 'L, RGB' not 'L + Synthetic L from RGB, RGB'

I'll review the notes in light of this discussion and change or clarify where needed. I need to make clearer where the techniques do not produce documentary results.

I will also go through the 'Special Techniques' section and see if I can make things clearer regarding 'Channel Interpolation' and 'Luminance, Color' settings.

I find it easy to get drawn in to looking for scientific rationale in techniques which are essentially artistic. The techniques which produce non-documentary results are open to modification according to the astro-artists' taste. There may be some scientific rationale behind a technique - like keeping the colour balance consistent but changing the luminance to change the balance of the composition.

This discussion has shown me once again how powerful and flexible StarTools is. - How many different ways there are to get good documentary results - and how many more ways there are to get artistic results.

I'm always trying to improve the User Notes so please drop me a PM if something is unclear (or wrong ).

Guy

[Mike in Rancho]

Yes, Guy, that would be good stuff to know, as well as why and to what extent documentary nature is lost or, perhaps nudged a little lol.

I do see where the composition came from now in that thread. Interesting on a couple levels. One is that I did my Cocoon nebula last month the wrong way, yet it turned out decent-looking, I think. Perhaps by luck. I will have to look at my logs to check on the exact way I extracted Ha and R and then combined. May have been a bit of a Frankenstein monster.

Second, that thread does indeed instruct throwing the OSC file into L, even though as noted above it does not properly take account of 2xG? Hmmm.

Thanks for your work, Guy!

[happy-kat]

I loved this post by Ivo and have used this technique several times since, I like the control it gives by being a blend at the latter end of processing an image. I will keep this approach in my bag of ideas even when using 1.8+ though I'll need to play more with 1.8 and the new NB accent to get familiar with it.
viewtopic.php?f=4&t=2148&hilit=m33

[Guy]

Thanks for reminding me of this - and also the improved control gained by using the Layer Module towards the end of the processing to blend the images.
So that makes three different approaches to combine narrowband data with visual spectrum datasets :- add it to luminance; add it to R, G, or B as appropriate; or use it as an accent.
Within those approaches there are also many different options with StarTools - and since we are not trying for a documentary result we are just limited by our own creativity - with some stunning results.

[Mike in Rancho]

I don't know if we are necessarily throwing documentary out the window. Depends what is done and how it is explained, of course.

And that's a good reason for a better understanding of how compose works with the luminance and chrominance, so that one knows what is doing, and how to describe what was done. Or of course we just say LLHaROIIIGB. That works.

Now it may not be visual spectrum accurate anymore, if L or certain channels are pumped up with NB, or both, but I think false or altered color is in the nature of narrowed emission bands. Just as long as it's not selective fake color procedures, right?

And if it's explained, an NB image with RGB stars is still mostly of a documentary nature - Both are real in their own way, just in a combination because we might find it to also be a little cooler aesthetically.