Separating SII and Ha

[Mike in Rancho]

I currently have a little bit of time on the Flaming Star and part of a nearby friend - framed in just to add a little spice. After two nights, I have amassed a huge 90m or so per NB filter or about 4.5 hours. So, a bit weak. Ha is predictably strong, SII isn't bad at all, but other than near the tadpoles, OIII is nothing but a haze and a blob.

Mapping and throttling to reveal SII in a way I like, however, is proving difficult - even though it's there.

Here's one of several renditions I managed. Don't mind the noise, it's the data plus some of these were done pretty quickly to get to color and see what I have going on.

IC405 2n SHO pstSVDb ST8 1C.jpg (480.96 KiB) Viewed 4704 times

This matrix was #3, 50/50 - 50/50 - 100, with 2.9 green reduce and 1.72 blue reduce. I guess there's a bit of green tint towards the Ha and perhaps red-orange tint to the SII, but mostly hard to tell.

I tested out a quick bicolor HSS (only loaded SII once), picked what amounts to H (H+S) S, with a red/Ha reduce of 3.0, which my understanding is a linear 1/3 multiplier, and came up with this. Nothing else here other than a big bin and then synthetic in SVD since I didn't have enough resolution left.

IC405 HSS test ST8 1B 1200.jpg (477.4 KiB) Viewed 4704 times

The bicolor separates Ha and SII quite nicely, I think. So, is it just because things get more mushed up when adding another channel to go tricolor? Or my settings? Or the pre-fab mappings, even though we have quite a few. It does seem to me that trying to reveal SII as red, when red is often a component of the Ha also due to mapping, could be an issue as well. And the Ha and SII regions do overlap quite a lot, but yet again they can be separated in the bicolor.

What would seem pretty cool would be to have that same brown-orange for Ha as in the bicolor, but have the SII reveal in bright, deep, or full red, and then also have my blue up in the corner. Assuming that the blue/OIII in and of itself isn't messing things up globally...

Seemingly, at least at full or pure intensities, the matrix #2, 40/60 - 40/60 - 100, ought to do this. Or perhaps #7, 50/50 - 85/15 - 100. #7 seems a bit stronger in the red rather than a more faded pastel. So says the little color mapping chart I had made up for all the matrices, anyway. But I get no luck with those, in fact they turn out even worse as the Ha and SII seem to blend together into a single solid hue.

Throwing a Hail Mary (that's for Dietmar to look up ) I tried an HOO bicolor, mapped as H(H+O)O, with SII NB Accenting, and....well, maybe let's not talk about that image.

If anyone wants to try out the specific data in question, the stacks can be found here: https://drive.google.com/drive/folders/ ... sp=sharing

I should be out there gathering more photons on this before the moon comes back, but, an evening of spiked eggnog seemed a better idea.

[admin]

I think your SHO rendition is pretty close to what I would expect for this object, and is very close to other renditions (renditions that I trust that is) I have seen. In this object, S-II and Ha correlate strongly, hence lots of yellow/orange (R + G). It's IMHO a high fidelity image.

Because of the strong correlation, as you surmise, a straight SHO->RGB palette (without any matrix magic) should maximize visibility of the three emissions (I had a quick go with the dataset, mapping straight SHO->RGB and it certainly does!

StarTools_2849.jpg (367.06 KiB) Viewed 4654 times

(sorry for the slightly different ST 9 German UI, but you'll get the idea )

It's one of those cases where having green in your image is definitely desirable, if you want to maximize the documentary worth of your image.

You are also right that by assigning the three emission 3 equal parts of the spectrum, but with O-III just being "wasted" on one part, a bi-color may indeed make things a more discernible. An H(S+H)S or S(S+H)H composite is indeed a great way to go. Most screens/color spaces (and human eye balls) are able to render and detect variations in green the best, variations in red second best and variations in blue poorly.

ST's separation of luminance and chrominance tries to take this into account in the Color module, but there is only so much intensity compensation you can do (for example the blue channel of a screens typically maxes out at just 20% of the max intensity of the green channel!).

PS spiked eggnog is always a good idea.

[Mike in Rancho]

Thanks Ivo! Appreciate the time and your tips.

I will try that out. I like the separation there that wasn't really revealing itself with a matrix choice, but will have to see if the big Ha reduction holds up on screen at my chosen resolution, especially with limited integration. I've noticed that can give a sense of noise, though OIII here is also reduced which could help.

I'll often take a quick peek at a no matrix straight SHO mapping, but usually end up with one of the blends. In fact one where my final was straight SHO was actually the Veil. That seemed to have physical separation of emissions, leading to a more striking image at full R, G, and B. Interesting that it works well here too with the strongly correlated emission regions.

I think I also shied away from it here, however, due to a heavy resultant purple cast.

Which leads to the obvious 1.9 inquiries.

I like the renamed channel bias legends. But I also notice the Cap green slider has been replaced with a Violett begrenzen slider. Or, shieberegler? Either way, I don't have one! I see you've also maxed it to 100%.

I'm thinking there's either a green/magenta toggle I can't see, or the cap swaps itself depending on broadband or narrowband.

I can make do for now, and in fact for this very target have already once employed a Layer/Color invert green cap, though it was with a matrix blend. But that raises questions about the documentary effect of green capping and well, how it works.

I'm figuring the cap (green going to yellow or brown) is the last function applied. First channel bias, then mapping, then capping? At least if capping has some relation to Max RGB and is just altering the very remnant green...or purple. If it's the other way around, however, say channel bias, then capping (or maybe conversion to another full color), then mapping, all should still be good I would think? Though that would change any kind of predictive charting of SII, Ha, and OIII mapping blends by punching the RGB numbers into a color generator.

[admin]

Thanks Ivo! Appreciate the time and your tips.

I will try that out. I like the separation there that wasn't really revealing itself with a matrix choice, but will have to see if the big Ha reduction holds up on screen at my chosen resolution, especially with limited integration. I've noticed that can give a sense of noise, though OIII here is also reduced which could help.

Denoise did a great job for me (though that too has been tweaked for 1.9)

I'll often take a quick peek at a no matrix straight SHO mapping, but usually end up with one of the blends. In fact one where my final was straight SHO was actually the Veil. That seemed to have physical separation of emissions, leading to a more striking image at full R, G, and B. Interesting that it works well here too with the strongly correlated emission regions.

I think I also shied away from it here, however, due to a heavy resultant purple cast.

Which leads to the obvious 1.9 inquiries.

I like the renamed channel bias legends. But I also notice the Cap green slider has been replaced with a Violett begrenzen slider. Or, shieberegler? Either way, I don't have one! I see you've also maxed it to 100%.

I'm thinking there's either a green/magenta toggle I can't see, or the cap swaps itself depending on broadband or narrowband.

I can make do for now, and in fact for this very target have already once employed a Layer/Color invert green cap, though it was with a matrix blend. But that raises questions about the documentary effect of green capping and well, how it works.

Yep, it implements exactly that technique (Layer->Inverse->Color->Cap Green->Layer->Inverse) straight in the module, because Cap Green is not useful for a SHO image.

The premise behind cap green (or cap purple for that matter), is that dominance of a hue can be argued to be indicative of an issue with the dataset, or an impediment to the goal of the observation.

For Cap Green the reasoning is that green channel dominance in outer space is exceedingly rare in the visual spectrum (notable exception being pure O-III emissions). Following that, any green dominance, must be the result of some unwanted influence (chromatic noise, a gradient remnants, etc.). Following that, it should be acceptable to truncate the green channel's level to that of the maximum of the other channels (red or blue, whichever is greater). It's a fair few assumptions, and with all the other pre-processing steps and tools in StarTools, using Cap Green should almost never be needed.

For Cap Purple, the reasoning is that purple dominance is exceedingly rare in a SHO composite; Ha (green) is almost always much stronger than S-II or O-III. Therefore if both S-II and O-III dominate Ha, we can assume that some sort of signal aberration is taking place. One place where such an aberration indeed usually exclusively takes place is the stars; they emit little in Ha, but much more in S-II (red) and O-III (blue). Red + blue makes purple. This is why stars often appear purple in SHO composites. In this case, star color really doesn't offer much in the way of scientifically useful cues other than "yep, stars still don't emit much in Ha today", and it can be argued they are somewhat distracting from more useful (and subtle) color cues.

Modules in StarTools 1.9 now adapt to the dataset or object your are processing, invoking more useful presets or adapting the UI (or even functionality) to match the use case better. That's why the UI in the screenshot looks different now, as StarTools is now aware of what type of data you are processing. For that purpose, the Compose module now has two selectors to better define the type of dataset and type of object(s) you are processing.

Let me know if you'd like a 1.9 preview (in English )

[Mike in Rancho]

Thanks Ivo, a sneak preview would be great!

In broadband the cap does seem to make good logical sense, following appropriate star sampling and correction, if needed. And I'm thinking that your description of the green cap, being a drop to the level of the highest other channel, may explain why capping seems able to result in two extra colors in MaxRGB - yellow or cyan. I've always wondered about that.

I need to think it through more for narrowband SHO, however. Choice of matrix and throttling amounts could make use of capping a bit complicated. While the AP consensus holds purple as unfit for human consumption, I want to make sure it doesn't have real meaning.

Straight SHO may be a safer application of capping though. Setting aside the stars, a more widespread purple could perhaps be more a result of noise or background levels picked up by the OIII filter? They seem to be prone to it. But that's what I was seeing with the Flaming Star here, and so that may not be a very legit OIII component to be purpling up all that outer space.

[admin]

Thanks Ivo, a sneak preview would be great!

I just uploaded a 1.9 preview version to the download section (see also this post).

In broadband the cap does seem to make good logical sense, following appropriate star sampling and correction, if needed. And I'm thinking that your description of the green cap, being a drop to the level of the highest other channel, may explain why capping seems able to result in two extra colors in MaxRGB - yellow or cyan. I've always wondered about that.

That's exactly what's going on!

I need to think it through more for narrowband SHO, however. Choice of matrix and throttling amounts could make use of capping a bit complicated. While the AP consensus holds purple as unfit for human consumption, I want to make sure it doesn't have real meaning.

Good point actually. Really, this whole "cap" business should take place before matrix translation...

Straight SHO may be a safer application of capping though. Setting aside the stars, a more widespread purple could perhaps be more a result of noise or background levels picked up by the OIII filter? They seem to be prone to it. But that's what I was seeing with the Flaming Star here, and so that may not be a very legit OIII component to be purpling up all that outer space.

Depending on influences (e.g. light pollution), that is a possibility. Though separation of luminance and color should hopefully reduce luminance vs color correlation, so that noise in one particular band does not necessarily show up once color is introduced.

[Mike in Rancho]

Thank you Ivo!

For the "final" image post on Nov 30, this is what I came up with. A bit tough to do the purple capping later on, as it seems one must get out of tracking to keep the channel balances and other settings from implementing again. So you can't quite predict what you are going to end up with until you've gone through a few iterations matching a log or via STReplay.

IC405 2n SHO ST8 5C.jpg (477.18 KiB) Viewed 4490 times

Then last night I took a quick run at it with 1.9 preview, mostly defaults, and setting a few options that I don't understand yet , and pretty rapidly came up with a version that I thought was better overall than from 1.8.

IC405 2n SHO ST9 1B.jpg (469.51 KiB) Viewed 4490 times

Some base processing could be different - I think the Contrast settings may have had different locality, and I probably needed to back off a few things in 1.9 that came on a little strong (although improved) like SVD and DN when keeping settings at default or where they had been in 1.8. I'll try again when I can play with it more, and I think keeping more resolution for SVD or reducing iterations would help prevent squaring off of small stars, and I'll also think about smaller grain for DN.

Rather solid though for an alpha preview, no crashes or hangups. Just a matter of learning the new toys and how to take advantage of them.