StarTools

These images and text were uploaded to Astrobin this morning. NGC 1955: An LMC Nebula Revealed

This is an experiment, so comments are welcome!

***************

Actually, NGC 1955 is the name of the open cluster associated with the emission nebula in this image. They are located in the Large Magellanic Cloud. This is a good three channel narrowband target, with strong Ha and OIII signals and a useful SII signal. 

This is my second image in an experiment inspired by the writing of Astronomer Travis Rector, author of Coloring the Universe and related scholarly articles. Professor Rector argues that processors of emission nebulas should try to differentiate among a nebula’s ionized gases, while at the same time being mindful of color design theory.

I used PixInsight to study the six palette options available for three channel narrowband images. These images are just simple channel combinations with temporary screen stretches--they are primitive, but useful.  

The first graphic below shows the three narrowband masters. Ha-top left, OIII-top right, and SII-bottom. The second graphic below shows the PixInsight channel combinations. Top row, left to right: SHO, OHS, OSH. Bottom Row, left to right: HSO, SOHl, HOS.

These six comparisons do not include the typical "Hubble Palette" where the green hue in an SHO palette has been substantially shifted toward red. I do not know how to reliably construct a color key for that situation. Furthermore, shifting green means that the color space is not being fully sampled. 

I felt that OSH might be the best tradeoff between scientific information and aesthetic appeal, but your preference may differ.  

Of course, assigning OIII to the red channel and Ha to the blue channel is not conventional, but the image manages to document the distribution of the three ionized gases, and also results in a pleasing color design, approximately using red-orange as the key color plus blue-violet and yellow-green as triadic complementary colors. (The names for these hues are taken from my color wheel.) 

Tech Notes for ASA 500/3.6:

ASA Newtonian, 500 mm aperture, 1900mm focal length, F3.6
FLI Proline 16803, 9 μm pixel, 4096 X 4096
ASA DDM85 equatorial mount
Processing with PixInsight, StarTools, and Affinity Photo

I think I commented about this elsewhere, but there are good reasons to tone down green.
As I wrote elsewhere too, colorwheel theory is indeed behind most of the hues we pick to show narrowband most effectively, however this is only part of the story;

We have to also take into account the rendering media's ability to render the different colors and - very important - human sensitivity to different wavelengths;
(from https://www.researchgate.net/figure/Sen ... like so;
As with all things, the resulting perceived brightness (Y) is an approximation, but it illustrates the required ballpark disproportionate weighting of the different colors quite well.

Crucially, for the purpose of this discussion, it demonstrates that a small variation in green yields a disproportionally large change in perceived detai (one of the reasons why a Bayer matrix takes two green samples versus just 1 sample for blue and red - we're much better at seeing green noise). Conversely, enormous amounts of blue channel changes are needed to show that same amount of perceived detail change.

In other words, pure green is terribly overpowering in a scene. So, if you wish to use coloring to draw people's attention to specific details in specific wavelengths, you should be using green sparingly compared to the other three color channels.

This is one of the major reasons why people steer away from using green in renditions, or - at the very least - try to take into account the actual perceived brightness different wavelengths cause. This is also the major reason why StarTools processes luminance and chrominance separately - it allows you to process detail regardless of coloring, while - at the very end - making sure that coloring is adjusted for perceived detail.

Mapping, say, SHO straight to RGB and calling it a day, though well-meaning, does not take into account the perceived brightness of different colors, and therefore will create images that are much harder to interpret for humans.

Mapping perceived colors to perceived brightness is a whole other can of worms (ST relies on CIELAB, but it is not a silver bullet either).

Hope this helps!

Hi Ivo,

I was delighted to read your comments. Until now, I haven't been able to find any useful advice on how to achieve Rector's recommendations: balancing scientific accuracy with aesthetic appeal. In fact, among amateur imagers I've found zero interest in accurate documentation of the chemical composition of emission nebulas. I've been groping for solutions on my own.

I guess the core question is whether Rector's point of view is really achievable. What if the hue of green were kept the same, but luminosity (and/or saturation) reduced? Or the hue of blue kept the same, but luminosity (and/or saturation) boosted? Could changes like those improve aesthetics while preserving the hue-based "mapping" of the ionized gases?

I'm an experimental person by nature, but sometimes that leads me into water that's way too deep!

Russ

PS I might add that I live in Southern Arizona, near Kitt Peak. Travis Rector worked as an astronomer at Kitt Peak for many years. In these parts, he is still a venerated person.

Russ.Carpenter wrote: ↑Sun Feb 06, 2022 11:26 pm I guess the core question is whether Rector's point of view is really achievable. What if the hue of green were kept the same, but luminosity (and/or saturation) reduced? Or the hue of blue kept the same, but luminosity (and/or saturation) boosted? Could changes like those improve aesthetics while preserving the hue-based "mapping" of the ionized gases?

That is precisely what StarTools aims to achieve in the Color module and the engine's strict separation of luminance and chrominance signals;
E.g. process a SHO composite as normal (import S:H:O as R:G:B, and have StarTools create a synthetic luminance all through the default settings). Process the detail to your liking. Once opening the Color module, SHO for the coloring is mapped precisely to RGB by default, taking into account perceived brightness of red, green and blue while doing the color mapping; brightness is dictated by the luminance (detail you processed), while hue is dictated by the chrominance.

It's somewhat hard to see (on purpose!), but if you use a dropper, you should notice green being subdued, while blue is being artificially boosted by "borrowing" brightness from the other channels (blue doesn't have enough "oomph" on its own).

This achieves the best of both worlds; informative coloring, without coloring impacting the detail that can be perceived.

I just happened on a test image I used when developing the Color module years ago. It neatly demonstrates how StarTools treats luminance and color, in order to coloring from perceptually influencing brightness.

Take this test image; If your monitor is properly calibrated (and of reasonable quality), the three red, green and blue squares in the last column should appear of disparate brightness, with green dominating (even though the squares max out the red, green and blue channels!).

You can import this image into the compose module as red, green and blue and let ST create a synthetic luminance dataset.
Then, going straight into the Color module and setting all ratios to 1.0, setting stauration too 100% and setting bright/dark saturation to full, this is how StarTools renders the coloring; If your monitor is properly calibrated (and of reasonable quality), the three red, green and blue squares in the last column should perceptually appear of equal brightness, rather than green dominating. All hues should be intact however (as far as gamut allows). Note that the quality and proper calibration of your monitor plays a very big role here (one of the reasons why this is so important!).