Help needed with narrow band processing

[dariv]

I have some data I took of M17 in narrow band, and pre-processed the data.
I have spent hours trying to get an acceptable final image in Startools. I don't know what I'm doing wrong or perhaps the data isn't very good?
I've tried several methods for combining the data but end up with purple background, or red stars, or not much color at all.
I'd appreciate it if someone could please take a go at the data and see if there is a decent image hiding in there!

Ha 258 minutes https://www.dropbox.com/s/tzq62nnskem0d ... .fits?dl=0
Oiii 123 minutes https://www.dropbox.com/s/ivbsihke6gi0q ... .fits?dl=0
Sii 147 minutes https://www.dropbox.com/s/q7p7fp9apickc ... .fits?dl=0

Thanks!

[admin]

Hi,

There are many, many ways you can do narrowband processing. You have a lot of artistic license when it comes to coloring.
Given you have acquired SII, H-alpha and O-III data, I'm just going to assume you are going for that classic Hubble look (do let me know if that isn't the case!).

In ST, you can import S II, H-alpha and O III in the LRGB module as red, green and blue respectively. You're then ready to process.

When processing narrowband data, it's usually a good idea to be gentle with the Wipe module (gradients and light pollution in narrow-band are much less of an issue). Usually, You'll want to use Wipe only for some simple bias removal. You may even wish to set Aggressiveness to 0 (this will only remove global bias, but will not model & remove any latent gradient). In an upcoming version of StarTools there will be a "NrwBand" preset for exactly this purpose.

Process detail as usual.

When you make it to the Color module, for that classic 90s Hubble look, you'll want to use Artistic, Not Detail Aware as Style, a low Bright Saturation (say 1.0-2.0) and a tweaked Dark Saturation (1.0-2.0) as needed, with much higher Saturation Amount. You will also want to cange LRGB Method Emulation to RGB Ratio, CIELab Luminance Retention or 50/50 Layering, CIELab Luminance Retention
This emulates the more primitive tools astrophotographers had at their disposal in that era (e.g. it liberally stretches and squashes color information along with the stretching of the detail). In an upcoming version of StarTools there will be a "Hubble" preset to accomplish just that.

In terms of color balance, every reasoning goes out the window and there are no set rules. Just pull on the Color Module's Red/Green/Blue Bias Reduce until you've found a color rendition you are happy with. Remember that the SHO Hubble palette is mainly meant to showcase the differences in physics/chemistry processes going on in different areas. Also be aware that some objects just don't have much signal in one or more bands (M17 is not that crazy rich in O-III for example, so blue will be lacking a little).

For inspiration for color ratios/schemes you can look at what others have come up with over the years for your object, but note that StarTools (like most other AP-specific tools) will only allow you to linearly scale the contribution to the red, green and blue channels of S, H and O respectively; it will not allow you to separately non-linearly stretch color channels, as that is not scientifically correct (use Photoshop or The Gimp afterwards if you really want to do that).

NASA images - especially from that era - also tend to be processed very conservatively (due to more primitive tools and focus on documentary value). So if you're going for that authentic 90s look, be conservative (or refrain from) local contrast and detail enhancement (except Decon).

If stars develop halos that are really too unsightly (quite common in SHO palette), use the Fringe Killer Filter Mode (typically used to remove halos caused by chromatic aberration) in the Filter module.

I processed your data as follows in (StarTools 1.5);

--- LRGB
Import S, H, O as R, G, B
--- Auto Develop
To see what we got.
--- Bin
Parameter [Scale] set to [(scale/noise reduction 50.00%)/(400.00%)/(+2.00 bits)]
--- Crop
Parameter [X1] set to [102 pixels]
Parameter [Y1] set to [36 pixels]
Parameter [X2] set to [2311 pixels (-90)]
Parameter [Y2] set to [1749 pixels (-45)]
--- Wipe
Parameter [Mode] set to [Correct Color & Brightness]
Parameter [Precision] set to [128 x 128 pixels]
Parameter [Dark Anomaly Filter] set to [6 pixels]
Parameter [Drop Off Point] set to [100 %]
Parameter [Corner Aggressiveness] set to [100 %]
Parameter [Aggressiveness] set to [0 %]
--- Auto Develop
Parameter [RoI X1] set to [314 pixels]
Parameter [RoI Y1] set to [463 pixels]
Parameter [RoI X2] set to [1519 pixels (-690)]
Parameter [RoI Y2] set to [1247 pixels (-466)]
--- HDR
(just to tame the core a bit - note that NASA would likely not do this!)
Parameter [Small Detail Precision] set to [Max]
Parameter [Channels] set to [Brightness Only]
Parameter [Algorithm] set to [Tame Highlights]
Parameter [Dark/Bright Response] set to [Full]
Parameter [Detail Size Range] set to [1000 pixels]
Parameter [Strength] set to [1.2]
--- Deconvolution
Default parameters.
--- Color
Parameter [Bias Slider Mode] set to [Sliders Reduce Color Bias]
Parameter [Style] set to [Artistic, Detail Aware] (instead of Artistic, Not Detail aware; cheating a bit here, as this wouldn't have been available in the 90s either)
Parameter [LRGB Method Emulation] set to [RGB Ratio, CIELab Luminance Retention]
Parameter [Dark Saturation] set to [2.00]
Parameter [Bright Saturation] set to [1.00]
Parameter [Saturation Amount] set to [300 %]
Parameter [Green Bias Reduce] set to [5.02]
Parameter [Red Bias Reduce] set to [1.47]
--- Wavelet De-Noise
Parameter [Smoothness] set to [80 %]

NewLRGBComposite.jpg (894.99 KiB) Viewed 10026 times

Hope this helps!

[dariv]

Ivo,
Thank you for the tips!
I have a few questions:
What made you select 6 pixels for the dark anomaly filter?
In the HDR module you went with max dark/bright and detail size of 1000 pixels, please explain?
Also, I've read on CN some people recommending binning as the last process, not one of the first. Reason being it produces a slightly cleaner image. Do you have a comment on this?

When initially combining the channels in the LRGB module, would it be better to adjust the weights for different exposure times? For example, if I have 2hrs of Ha and 1hr of Oiii, boost the Oiii contribution to 2?
And finally, is it better practice to process each channel separately and then combine into a final image?

Thanks!

[admin]

Ivo,
Thank you for the tips!

Glad I could help. Is this Hubble palette rendering what you are after?

I have a few questions:
What made you select 6 pixels for the dark anomaly filter?

Good question. It usually doesn't hurt to bump up this parameter in the case of noisier datasets and especially with the way I used Wipe here (e.g. simple bias subtraction, but not gradients). "Dark noise" can work the same as "dark anomalies" in that Wipe may back off. However because the dark noise amount is the same everywhere in the image, it won't create halos like local dark anomalies would. Instead it just causes Wipe to leave a tiny bit of bias in.

In the HDR module you went with max dark/bright and detail size of 1000 pixels, please explain?

"Dark/Bright Reponse" sets how strong the effect of the algorithm is in the brightest and darkest parts of the image. Leaving it low causes the effect to be mostly visible in the midtones, whereas setting it higher causes the effect to be visible in the shadows and highlights as well.
Detail Size range 1000px was chosen to better blend in the recovered detail with the rest of the image.
Have a look at the HDR module information on the website or Guy's excellent User Notes on HDR;

usage

Going beyond the presets, more detailed adjustments can be made, starting with the 'Detail Size Range' parameter. This parameter is highly influential on the end result. It governs the range of detail sizes HDR should concentrate on, in order to bring out the most detail. Keeping this value small will see small detail accentuated. However, using larger values will see both small and large structural detail modified. Using larger values will progressively dig out larger scale
structures and can be quite effective in highlighting these.

In order to throttle how much the shadows and highlights respond to the enhancements, a brightness mask is used, the power of which is controlled by the 'Dark/Bright Response' parameter.

Also, I've read on CN some people recommending binning as the last process, not one of the first. Reason being it produces a slightly cleaner image. Do you have a comment on this?

That sound rather curious. The whole idea behind binning is to obtain a dataset that is less noisy, so the algorithms you apply can do their work better. Doing it after just means you just did all your processing with worse signal than you could have had. This doesn't make sense to me at all...

When initially combining the channels in the LRGB module, would it be better to adjust the weights for different exposure times? For example, if I have 2hrs of Ha and 1hr of Oiii, boost the Oiii contribution to 2?

You could definitely consider doing that. In that case you'd want to create a synthetic luminance frame that combines/weighs S-II, Ha, O-III into one "black and white" (luminance) dataset. Then you process this dataset to the standard of detail you wish to show. You can then separately process a SHO->RGB stack, just for the colors (e.g. don't worry too much about the detail). When you're doing, you then combine the processed luminance and color image using the Layer module (Color of foreground/Brightness of foreground as Layer mode).[/quote]

Hope this helps!

[dariv]

Glad I could help. Is this Hubble palette rendering what you are after?

I don't really have a specific style in mind. Just whatever process looks the "best."

I'd say something clean with as much detail and pop as can be squeezed out of the data set.

[dariv]

Tried going with a synth lum and got this. not too bad, I'll keep at it....
https://www.dropbox.com/s/6j8ig471lq70k ... p.jpg?dl=0

[admin]

Totally subjective of course, but I love the copper coloring and detail!

[Rkonrad]

This is what I came up with.
https://www.flickr.com/gp/rkonrad/CXx45C

Given the HA concentration I think, for me, the overall predominance of colour should be yellow copper or green. My attempt was created by processing the 3 bands individually and the signal in each is arbitrary so theres not a lot of science here. In another attempt I processed them using dev at 90% and then blended them at the lrgb screen each according to the capture time. I for example I increased the Oiii ratio by 2 and Sii by 1.75 but the result was an overly and almost completely blue object. Not sure why it turned out like it did considering that I tried to equalize the exposures.....

Great data Dariv!

Cheers Richard

[dariv]

Richard,
That's very nice!
I've grabbed a lot more data, a total of about 30 hours; 10 on each filter.
BTW, I use APP to do all the pre-processing.
If anyone is interested, I can share the data.
Thanks again!
-DaRiv

[admin]

If anyone is interested, I can share the data.
-DaRiv

30h! I'd definitely be interested!
I could really use some good real-world test data for a feature I'm working on.