Workflow to increase OIII in duo band images

Questions and answers about processing in StarTools and how to accomplish certain tasks.
Post Reply
jlh
Posts: 23
Joined: Sat Nov 04, 2023 1:06 pm

Workflow to increase OIII in duo band images

Post by jlh »

Is there a suggested workflow to boost the OIII data and color in processing OSC images captured with an Ha/OIII duo band filter? Should the extracted Ha and OIII files be processed separately with the OIII signal stretched more (and/or the Ha less), and then combined in Layer in a way to prevent the Ha from overpowering the OIII? Or is it better to load them both in Compose and try to increase the OIII for the main object (without adversely affecting the colors/hues elsewhere) in certain modules? I can't get the OIII to come through satisfactorily using the suggested HOO loading in Compose and HOO processing in Color.

A second question is how would you deal with star colors in the context of that suggested workflow?

Thanks,
Jeff
Mike in Rancho
Posts: 1158
Joined: Sun Jun 20, 2021 10:05 pm
Location: Alta Loma, CA

Re: Workflow to increase OIII in duo band images

Post by Mike in Rancho »

Hi Jeff,

Do you have a particular dataset in question that isn't ending up how you think it should? Or just general targets in mind?

I know the subject has been discussed here often, directly and indirectly, but you might have to go hunt through threads in the gallery and image troubleshooting sections. Plenty of varying theories of what can be done, as well as what should be done (i.e. in terms of maintaining fidelity to the data or at least a good logic in presenting the data), too.

Some, maybe many, targets are just weak in OIII. In mono you will sometimes see people amass a much longer integration on OIII. That greater SNR permits a greater contribution to the luminance, though the relative chrominance remains the same. Not really possible with a duo though, which will have identical integration.

Using ST's typical Compose for duoband and bicolor matrices in Color, you will be held to relative Ha/OIII balancing as the logical way to display HOO narrowband. While still holding to that ideal, there's a couple things you can try. Throttling back the Ha, of course, to start. Toggle the MaxRGB screen to get an idea of overall Ha/OIII balance and if you can see stronger (with Ha relatively throttled) OIII regions start to reveal themselves. Of course this weakens the Ha, and in some targets can turn fringe areas into gray fog.

Try boosting the saturations, possibly by quite a lot, and balance the light and dark sats towards emphasizing the blue OIII, if possible. Reverting style back to Color Constancy may help pump the blue back up, also.

If a noisy OIII channel out of your duo is causing problems with this, it's possible some adjustments at the Compose stage can help. You can indeed extract the channels. ST can do it, but perhaps best done in your stacker to maintain 32-bit depth. Consider ditching the B channel entirely, depending on target, as it may be the weakest and noisiest. And if your duoband has an expanded OIII pass, such as an L-eNhance which also captures Hb, discarding B and only taking G as your OIII may remove pesky Hb (depending on the cutoffs of your Bayer filters). Then compose those two files into ST into R and G as normal R(GB)(GB) bicolor, or Ha into R and OIII into G and B but zero out the exposure time on either G or B.

The same thing may be possible using only Compose by not loading the B channel, and then picking normal R(GB)(GB) bicolor.

Depending on your estimate of the relative SNR in your channels, you can even make further adjustments using the exposure time sliders, which affect the creation of the Synth L. The idea behind the Synth L is the best overall SNR, and in narrowband that is not necessarily as well matched to exposure times as it would be in broadband RGB.

If you want to dismiss the need for relative emissions balancing, then yes you could extract and process separately as a grayscale, and see how a subsequent layer operation looks. However, it is a bit tricky to get the final results colorized for that later Layer usage, and not a good way to do it all on the fly, that I am aware of. But, I actually did something quite similar to this when I was processing some (six filter!) JWST data. After registration, each was processed separately in ST, and in Color I abused the sliders in order to get the hues I wanted. Step by step things were then blended in Layer, sometimes with different Layer operations (such as Lighten mode for one that I recall) depending on what was getting added in and where the data was spatially for layering in.

Of course that's taking a step (or three) towards "artist's impression," but I think NASA does the same thing, and hey, we don't have a six channel JWST compose module. Also, JWST is a mishmash of infrared wideband, narrowband, and medium band anyway. So the whole thing is far more about what can you show is where, spatially, with the colors, than it is relative emissions.

Whether you want to try a similar philosophy with bicolor Ha/OIII would be up to you.

More things you could try include a different channel mapping that may set off the blue better, perhaps H(H+O)O. Or even some of the tricolor SHO matrices. With data composed as R(GB)(GB), those mappings still end up as bicolors.

Finally as a bit of artistic boost you could see about an OIII boost in the Filter module. I think there's one in there, but I haven't used it in years.

Caveat though, things may still not appear similar to images of the same target you might find on the internet, with huge pools of incredibly deep blue OIII, yet at the same time incredibly strong red Ha everywhere else. Well there are a couple techniques for that, some of them quite popular, but probably best considered as airbrushing, IMO.

Star colors are a tough call. You can try to mostly bleach them to white, using increased highlight repair in Color, or color tame in Shrink if you use that (perhaps even as only deringing rather than iterations). A star mask and some Layer operations can also bleach the stars out, again if you'd rather they be white instead of the duoband, which can be odd-looking. Just the way most stars spit out Ha and OIII.

Another, and still logical, trick to try is capturing unfiltered OSC for the stars and blending those in, usually by a Color of Fg operation in Layer on a star mask, IIRC. Then you can say you've got duoband with RGB stars. A lot of posts here cover how to do that. I lack skill in pulling that off, at least where stars are in front of nebulosity.

Hopefully some of that helps? :confusion-shrug:

Others may be able to chime in with better suggestions. I know we've got a number of duboand users here. :D
jlh
Posts: 23
Joined: Sat Nov 04, 2023 1:06 pm

Re: Workflow to increase OIII in duo band images

Post by jlh »

Hi Mike,

Thanks so much for all of those suggestions. I will try various of them. Generally my problem seems to be trying to boost the OIII color vis-a-vis the Ha is not limited to the OIII signal but ends up adversely coloring the rest of the image. For what they are worth below is a link to some Ha/OIII files for the Soul and Heart nebulae. I know they are not great data and the Heart is kind of a clumsy mosaic. Kind of works-in-progress. FYI they were captured with an Optolong L-Ultimate filter and an ASI294MCP camera.

https://drive.google.com/drive/folders/ ... sp=sharing

Thanks again,
Jeff
Mike in Rancho
Posts: 1158
Joined: Sun Jun 20, 2021 10:05 pm
Location: Alta Loma, CA

Re: Workflow to increase OIII in duo band images

Post by Mike in Rancho »

Hi Jeff,

Ah the Heart and Soul. I was afraid of that. :lol:

These targets are infamous for Hey where's my blue?, and I nearly mentioned that in my post.

I'll take a look, but the share is still private, not global. In Drive you have to take it one step further in the menu and permit viewing by "anyone with link." Otherwise it would only be to specific named other Drive accounts you might share out to.
jlh
Posts: 23
Joined: Sat Nov 04, 2023 1:06 pm

Re: Workflow to increase OIII in duo band images

Post by jlh »

Hi Mike,
Sorry about that; try accessing now.
Jeff
Mike in Rancho
Posts: 1158
Joined: Sun Jun 20, 2021 10:05 pm
Location: Alta Loma, CA

Re: Workflow to increase OIII in duo band images

Post by Mike in Rancho »

Jeff,

That works! But as predicted, Heart and Soul just kick my behind. They always do. :(

Maybe Freddy can take a run at things? I believe he has extensive Soul experience... :D

What can you tell us about the integration numbers and sky conditions here? Looks like maybe 4.5h on Soul? The Heart mosaic I'm not sure how that all pans out. But either way I'm leaning towards more time needed for better SNR, which might permit more playing around.

A little more than a year ago I got around 3h on the Heart, 6-7nm SHO filters, B8, and it was...awful. :shock:

Even if some portions of these nebulae are bright, overall there's a good bit of faint structure that just needs amassing a lot of time, and I'd say that includes any OIII. And even then, much OIII is exactly overlain by stronger Ha no matter what.

How did you turn your duo stacks into the separate Ha/OIII files here? Mostly interested in how the OIII was constructed.

The Heart mosaic has a big anomaly in the OIII that really stood out in Wipe and also really didn't want to go away. I'm wondering if the way it was merged caused a strong gradient area to stitch up to a weak gradient area? Mosaics can be a headache that way, trying to ensure that you are stitching like-to-like as much as possible. :think:

After a few tries this is the "best" I got from the Soul data. By and large I composited and processed normally, but as we discussed the relative throttling of Ha to show off OIII also shows off the weakness of the SNR, disappeared Ha structure, and gray fogged the edges. So I started looking into any ways that NB Accent might allow some "boost" cheating - as long as we are good with tossing proper relative emissions out the window. It's really meant for augmenting full spectrum data with narrowband, but what good is a tool if you can't abuse it? :lol:

Trying to boost OIII as the NBA file just didn't work well. Way too weak and noisy of a file. So I started over but with Ha as the NBA file, and then when I got to color I heavily balanced towards the OIII. In the ensuing NBA, I then fed some Ha back in to make up for that, but not at 100% strength so as to not ruin whatever blue I had obtained earlier. I did use pure red setting.

It's kind of a goof, as the structure is identical to what was already loaded and processed, but I think does help with the gray fog in the outer reaches of the target.

If I were to do it again, I'd probably either pull back more on the Ha in Color beforehand, or lessen its saturation, because the end blending I got out of NBA may be bleeding a bit in some parts. And then I overcooked the denoise but didn't realize it until too late, and I didn't feel like starting over again from the beginning just to re-denoise it.

Any success in your further processing of it?

JLH Soul Test2 ST9.jpg
JLH Soul Test2 ST9.jpg (557.55 KiB) Viewed 3733 times
jlh
Posts: 23
Joined: Sat Nov 04, 2023 1:06 pm

Re: Workflow to increase OIII in duo band images

Post by jlh »

Hi Mike,

Wow; what great work using my data. You have great ideas for me to try in StarTools!! (Did you start with the Narrow Band settings in Wipe with one process after Compose? The OIII seems relatively very noisy which I understand usually happens).
Regarding my data; it was 61 300 second light frames captured with an ASI294MCP and an Optolong L-Ultimate filter and an AT65EDQ. The seeing was so-so from my Bortle 8+ front yard (actually about as good as it can be from that location; the guiding was around .75" for the night). I used APP to split out the Ha and OIII integrations. I can't remember exactly but I probably used some extra local normalization correction when integrating. Since I was using mostly the defaults I think I forgot to disable background neutralization when normalizing the light frames (as StarTools recommends).

Thanks again,
Jeff
Mike in Rancho
Posts: 1158
Joined: Sun Jun 20, 2021 10:05 pm
Location: Alta Loma, CA

Re: Workflow to increase OIII in duo band images

Post by Mike in Rancho »

Hi Jeff,

Thanks, I think it looks okay, at least for a test result. I actually did not use NB preset in Wipe. Anytime OIII is involved, and especially via OSC+duo, the gradient just overpowers that preset. So I left it at the default 75 and bumped the DAF.

I am unsure how APP does extractions like that. Possible it combines the B and G the way ST does in Compose for OSC/DSLR bicolor? If so, and I touched on this a little in my first post (somewhere in there, I know I threw everything but the kitchen sink at it :lol: ), is that I think in this case it might be good to drop the B entirely in favor of the G.

Looking at the graphs for the 294MCP and the L-Ultimate, that very thin band of 3nm OIII falls at over 90% QE (for whatever these graphs and numbers are worth) for G, but below 50% for B. So I surmise that inclusion of B here could degrade the OIII and make it look worse. :think:

If you have the original stacks, they can perhaps be manually extracted. PI, Siril, probably APP too should be able to do that readily enough in 32-bit FP.

Local normalization should be fine. I use that in PI-WBPP and there's a thread here on doing that. I didn't know APP had their own version. But, you probably want to check on what is/was meant by background neutralization. If that's anything like channel neutralizing (i.e. white balancing), it could have had negative consequences to this duo narrowband stack, which we've already mentioned will be unusually weak in B. So using B in any neutralizing I think could make a mess of things.

So who knows, perhaps a re-stack and then a different way of going about the extractions, and we may get a much better starting point of datasets to see what we can do with them.

FWIW, here's an edited log for the result above. Nothing to take as gospel here, I was experimenting on the fly and I'm sure made errors. As said I would totally re-do the denoise with lower grain, not as low scale correlation, or at least feed back in some equalized grain.

Code: Select all

-----------------------------------------------------------
StarTools 1.9.561beta
Wed Jan 10 2024
-----------------------------------------------------------

Loading red channel data
File loaded in LRGB module [G:\ASTRO\2-ST FOLDER\jlh\SoulNebulaHa.fits].
Loading green channel data
File loaded in LRGB module [G:\ASTRO\2-ST FOLDER\jlh\SoulNebulaOIII.fits].
Loading blue channel data
File loaded in LRGB module [G:\ASTRO\2-ST FOLDER\jlh\SoulNebulaOIII.fits].
Loading NB accent data
File loaded in LRGB module [G:\ASTRO\2-ST FOLDER\jlh\SoulNebulaHa.fits].

--- Compose
Parameter [Luminance, Color] set to [L + Synthetic L From R(2xG)B, R(GB)(GB) (Bi-Color from OSC/DSLR)]
Parameter [Color Ch. Interpolation] set to [On]
Parameter [NB Accents Type] set to [H-alpha/S-II from NB Filter]
Parameter [Spectrum and Filters] set to [Duoband/Triband/Quadband (OSC or DSLR)]
Parameter [Scene] set to [Field-filling Nebula]
Parameter [Total Exposure] set to [Not Set]
Parameter [Blue Total Exposure] set to [Not Set]
Parameter [Green Total Exposure] set to [4h35m (275m) (16500s)]
Parameter [Red Total Exposure] set to [4h35m (275m) (16500s)]
Image size is 4264 x 2961
Type of Data: Linear, was not Bayered, or was Bayered + white balanced

--- Crop
Parameter [X1] set to [84 px]
Parameter [Y1] set to [145 px]
Parameter [X2] set to [4180 px (-84)]
Parameter [Y2] set to [2816 px (-145)]
Image size is 4096 x 2671

--- Bin
Parameter [Mode] set to [Fractional]
Parameter [Scale] set to [Scale 50.00% / +2.00 bits / +1.00x SNR Improvement]
Image size is 2048 x 1335

--- Wipe
Parameter [Synthetic Dark/Bias] set to [Off]
Parameter [Gradient Edge Behavior] set to [Absorb 50%]
Parameter [Synthetic Flats] set to [Off]
Parameter [Sampling Precision] set to [256 x 256 px]
Parameter [Dark Anomaly Filter] set to [3 px]
Parameter [Gradient Falloff] set to [0 %]
Parameter [Synth. Bias Edge Area] set to [100 %]
Parameter [Gradient Aggressiveness] set to [75 %]
Parameter [Correlation Filtering] set to [Off]

--- RoI Optimized Develop
Parameter [Ignore Fine Detail] set to [7.0 px]
Parameter [Outside RoI Influence] set to [15 %]
Parameter [RoI X1] set to [458 px]
Parameter [RoI Y1] set to [262 px]
Parameter [RoI X2] set to [1571 px (-477)]
Parameter [RoI Y2] set to [1076 px (-259)]
Parameter [Detector Gamma] set to [1.00]
Parameter [Shadow Linearity] set to [25 %]

--- HDR
Parameter [Signal Flow] set to [Tracked]
Parameter [Quality] set to [Low]
Parameter [Gamma Shadows (Lift)] set to [1.00]
Parameter [Gamma Highlights (Tame)] set to [1.08]
Parameter [Gamma Smoothen] set to [20.0 px]
Parameter [Context Size] set to [30x30 pixels (1.46% image W, 2.25% image H)]
Parameter [Shadows Detail Boost] set to [30 %]
Parameter [Highlights Detail Boost] set to [15 %]

--- Spatially Variant PSF Deconvolution
Mask used (BASE64 PNG encoded)
[MASK CODE DELETED]
PSF samples used (8 PSF sample locations, BASE64 encoded)
VFMAAAgAAAg3BVYH4wPcBe4DXAXGBF8AFgUdAr8D9gZsAKwD/gAvAP4A
Parameter [Synthetic PSF Model] set to [Circle of Confusion (Optics Only)]
Parameter [Synthetic PSF Radius] set to [1.5 px]
Parameter [Synthetic Iterations] set to [Off]
Parameter [Spatial Error] set to [1.00]
Parameter [Deringing Focus] set to [50 %]
Parameter [Dyn. Range Extension] set to [1.00]
Parameter [Deringing Fuzz] set to [50 %]
Parameter [Sampled Iterations] set to [10x]
Parameter [Linearity] set to [100 %]

--- Color
Parameter [Bias Slider Mode] set to [Sliders Reduce Color Bias]
Parameter [Style] set to [Artistic, Detail Aware]
Parameter [LRGB Emulation] set to [RGB Ratio, CIELab Luminance Retention]
Parameter [Matrix] set to [HOO Duoband 100R,50G+50B,50G+50B]
Parameter [Dark Saturation] set to [6.0]
Parameter [Bright Saturation] set to [9.0]
Parameter [Saturation Amount] set to [400 %]
Parameter [O-III Reduce Bias] set to [1.00]
Parameter [O-III Reduce Bias] set to [1.00]
Parameter [H-alpha Reduce Bias] set to [3.51]
Parameter [Mask Fuzz] set to [1.0 px]
Parameter [Cap Green] set to [0 %]
Parameter [Highlight Repair] set to [8 px]

--- VISUAL SPECTRUM NARROWBAND ACCENT COMPOSITING
Parameter [Ignore Fine Detail] set to [4.6 px]
Parameter [Outside RoI Influence] set to [15 %]
Parameter [RoI X1] set to [458 px]
Parameter [RoI Y1] set to [262 px]
Parameter [RoI X2] set to [1571 px (-477)]
Parameter [RoI Y2] set to [1076 px (-259)]
Parameter [Threshold] set to [13 %]
Parameter [Shadow Linearity] set to [50 %]

--- Visual Spectrum Narrowband Accent Compositing
Parameter [Response Simulation] set to [H-alpha/S-II (pure Red)]
Parameter [Color Modify Input] set to [Luminance]
Parameter [UNKNOWN] set to [Luminance]
Parameter [Detail Size] set to [0 px]
Parameter [Brightness Correlation] set to [Off]
Parameter [Gamma] set to [1.00]
Parameter [Strength] set to [70 %]
Parameter [Band Balance] set to [50% / 50%]

--- Unified De-Noise
Parameter [Grain Size] set to [7.0 px]
Parameter [Walking Noise Size] set to [1.0 px]
Parameter [Walking Noise Angle] set to [0]

--- Unified De-Noise
Parameter [Scale 1] set to [99 %]
Parameter [Scale 2] set to [95 %]
Parameter [Scale 3] set to [95 %]
Parameter [Scale 4] set to [95 %]
Parameter [Scale 5] set to [75 %]
Parameter [Equalized Grain] set to [0 %]
Parameter [Scale Correlation] set to [20 %]
Parameter [Color Detail Loss] set to [50 %]
Parameter [Brightness Detail Loss] set to [50 %]
Parameter [Grain Dispersion] set to [7.0 px]

File saved [].

I don't know if it means anything, what with using NBA on an already Ha-inclusive image, and that NBA performs a second and different non-linear stretch. :confusion-shrug: It's possible it just gave a little push to the strongest Ha regions (as well as the weakest OIII, meaning the gray fog), or perhaps it's just too artistic now.
jlh
Posts: 23
Joined: Sat Nov 04, 2023 1:06 pm

Re: Workflow to increase OIII in duo band images

Post by jlh »

Hi Mike,

Another approach I discovered is to use the Filter module to bring out OIII color. (It is amazing to me the tools I am finding available in StarTools). I masked the area where OIII is expected and masked the stars out of that area. Probably could have tweaked that mask better and I only targeted an area in the larger lobe of the nebula for this run. In Filter I used the Saturate H-beta/OIII Mode and clicked on non-OIII colors per that Mode's instructions. It looks fairly natural to me, at least preliminary. I am sure that experienced processors like you could do better. What do you think about this approach to emphasize and bring out OIII in duo band images?

Jeff
SoulFilter.jpg
SoulFilter.jpg (421.09 KiB) Viewed 2692 times
Mike in Rancho
Posts: 1158
Joined: Sun Jun 20, 2021 10:05 pm
Location: Alta Loma, CA

Re: Workflow to increase OIII in duo band images

Post by Mike in Rancho »

Hi Jeff,

Well, I don't know. :think:

I like a lot of the way your Ha structure looks. As far as Filter goes, well I don't use it that much anymore. Or at all I suppose. Now back in the day I used it quite a bit, mostly trying to hammer my crummy images into something more pleasing. But that was kind of just painting over my problems.

Not that Filter didn't do the job. Some stuff is so bad you just have to use reject on it. :lol:

Primarily though I think it's handy for things like doublet blue bloat.

The troublesome part is "masking the area where OIII is expected" starts leaning a bit painty. Now, if you could just mask out the stars, like in a Sharpen protective mask, that would probably be fine, but may not produce the results you are looking for since the filter would otherwise be global, and then you are just throttling and fogging the Ha again.

If you like it and think it fits in naturally, of course, go for it.

Another thing you might consider, as long as we are just playing around, is to try SS in saturate mode. See if that bumps up the appearance of your OIII. Modify gamma, saturation, and strength as needed, and if the red Ha starts bleeding too much, perhaps switch the Detail Preservation to Min Distance to 1/2 Unity. :confusion-shrug:

I tried that on your finished data I had from before (open non-linear no tracking) and kind of liked the blue boost. It did fog up the background too much, but I dialed that back by next going into FilmDev and dropping gamma 10%. One could also adjust any oversaturated red there too, by carefully lowering red luminance contribution. As long as we're just playing, of course. :D

Duoband bicolors can be tough, especially certain targets, and not having that third channel to offload things a bit from the pink/teal see-saw.
Post Reply