Units of bias sliders in Color?

[Mike in Rancho]

I tend not to get answers to these types of things, , but what the heck might as well try anyway.

Does anyone know if the numbers for the color bias controls (1 to 20 I think, to two decimals) mean anything? For example I did a bicolor the other night and I think I reduced OIII at a setting of 1.30. I don't think that would be a 30% reduction in OIII contribution, but, 3%, or something else, or...nothing that's an actual unit or percentage?

I did figure out that, in bicolor, the sliders are additive for the combined B and G. Thus, B reduction of 1.20 gives the same result as G reduction of 1.10 and B reduction of 1.10.

Anyway, just wondering (again). This is more just curiousity about how to describe the alteration being made (for example to non-ST users), and is not as objectively applicable as my prior attempt to find out what the units for SS Airy disk mean.

[admin]

I tend not to get answers to these types of things, , but what the heck might as well try anyway.

Does anyone know if the numbers for the color bias controls (1 to 20 I think, to two decimals) mean anything? For example I did a bicolor the other night and I think I reduced OIII at a setting of 1.30. I don't think that would be a 30% reduction in OIII contribution, but, 3%, or something else, or...nothing that's an actual unit or percentage?

I did figure out that, in bicolor, the sliders are additive for the combined B and G. Thus, B reduction of 1.20 gives the same result as G reduction of 1.10 and B reduction of 1.10.

Anyway, just wondering (again). This is more just curiousity about how to describe the alteration being made (for example to non-ST users), and is not as objectively applicable as my prior attempt to find out what the units for SS Airy disk mean.

Always happy to explain any and all parameters in as much detail as you need!

The Bias controls are simple multiplication factors for the different channels as imported (e.g. before any remapping with the Matrix parameter). The inverse (e.g 1.0 divided by the number) of the multiplication factors are taken, in case of the bias reduction mode.

E.g. when treating an RGB image, they simply multiply the red, green or blue channels (crucially, however, this is done in the linear domain!). When doing this for RGB images, this is also known as white balancing.

Similarly, when processing a HOO dataset (meaning HOO mapped to RGB), you can control H with red and O-III with green OR blue (e.g. both green and blue sliders have the same effect of throttling O-III, which will appear cyan).

As for the Super Structure module's Airy Disc Radius parameter, this governs how "big" the Airy Disc Point Spread Function is, by which the light in the image will be diffracted.

The "right" size is a function of the angular size of the image (e.g. the field of view) and the characteristics of the optical system (aperture, focal length). By looking at the Synth module, you can get a better idea of what goes into artificial PSF synthesis (and in screen #2, what happens when you apply said PSF) .

Given the subtleness of the effect of this parameter, and the many parameters you could apply to the mix, I opted for just a single, easy to control parameter that lets you "eyeball" something that is aesthetically pleasing and roughly matches the diffraction of your optical system + atmosphere for your dataset.

The whole reason for giving the Super Structure module at least some sort of an idea of how light was diffracted in your image, is so that it can better hide any unwanted effects in the smaller scale detail, while more successfully manipulating the very large scale detail (the latter being its purpose).

Hope that helps!

[Mike in Rancho]

Thanks Ivo, appreciate the time as it is clear you've been busy! And most of the time when a post doesn't get traction I just figure I probably asked a really dumb question.

Great, so the slider values are linear multipliers (or dividers). I am going to have to look up what that means, and it will probably force me to dig up a little RGB pixel theory too. I'm probably forgetting that, aren't I? Each pixel will have it's own R, G, and B value, and likely not 8-bit 0-255 either.

I will have to try out the Synth module for Airy disk training! I have generally avoided it as one of the heavily-artistic modules. And now that I have a Newt, spikes are free! But on CN it was recently mentioned there was star color in Synth also, so I thought I would look into it. Alas, the module description page was kind of a "coming soon" deal. So I'll read the manual or the user notes before diving in.

I have noticed the mere subtle changes of the SS parameter but, me being me, I always would like it to be the "correctest" setting possible other than just aesthetic appeal.

I don't suppose SVD can flash an average or center-image PSF on the screen for us after calculating the model, that we could then use somehow later? Like, ST has determined that your PSF is two.

But for now I'll check out Synth. Thanks!

[admin]

Thanks Ivo, appreciate the time as it is clear you've been busy! And most of the time when a post doesn't get traction I just figure I probably asked a really dumb question.

No such thing as a dumb question!

Great, so the slider values are linear multipliers (or dividers). I am going to have to look up what that means, and it will probably force me to dig up a little RGB pixel theory too. I'm probably forgetting that, aren't I? Each pixel will have it's own R, G, and B value, and likely not 8-bit 0-255 either.

Say your red bias is 1.3 and some 8-bit red channel pixel value is 55, then the new red channel value becomes 55*1.3 = 71.5. That's really all there is to it.

The important thing is that this is all done in the linear domain, so the stretched brightness/detail (luminance) is applied on top of the coloring afterwards.

I will have to try out the Synth module for Airy disk training! I have generally avoided it as one of the heavily-artistic modules.

While it can indeed make pretty spikes, it's ultimately meant as an educational tool. Rather than "painting things on", it actually goes through the whole process of modelling the optical train, after which it diffracts selected point lights and - yes - composites this onto the existing image.

One of the things I created it for, is to show people why stellar profiles look the way they look; the amount of times you still see people describing diffraction patterns as "halos" is too damn high! They are a natural thing, inherent to optical trains, and they can tell you so much about the optical train itself, as well as provide you with extra information on stars (the rainbow patterns).

[Mike in Rancho]

Thanks again Ivo. That is interesting.

So each channel is really adjusted independently. But I presume it is the 3D nature of each pixel that makes it appear that the effect is relativistic. Well, it guess it actually truly is relativistic the way the three channels create final pixel colors.

And the fact the multiplier is applied at a linear data level, which we can assume is a pretty low absolute amount, is why one can have multipliers like 20.0 at the max. Though both of those facts would make me start wondering about quantization. Must think more.

I played with Synth some but am somewhat lost, so more work to do there. I was able to put in all the correct parameters for aperture, focal length, and on page two the proper field of view, but it still seemed pretty mask dependent and while I could get a pretty close match, it wasn't spot on. For some stars the synth spikes would be much greater than reality, for others the opposite. And as far as the star itself, well, I'm a bit uncertain there. Might need to play more with brightness, gamma, and so on. Also not sure where I can find what it is showing me about the PSF that might be useful for how to set the Airy disk in SS.

Some of it could be that I was loading in a fully completed image, which probably had things like SVD, SS, and maybe even Shrink though I don't use that much except for deringing, already applied.

When I could get all the settings close though, Synth did seem kind of handy itself almost as a de-ringer, with the added capacity to return some proper original color to stars that had maybe become washed out a bit in processing. Interesting.

I maybe use the term halo all the time myself. Should it be limited instead to just artifacts of filters and the like? But if it is just the "normal" halo, even if using the term halo, I still have come around to understanding them as part and parcel of the optics and what was legitimately captured. And is one reason I raise my eyebrow at too much halo-fixing or goodness, star removal. The halo pixels were likely burned out somewhat by the stars, it's set into the data that way, so not like one can remove the halo and now see what was behind it.

However, even things like the ST module description for Shrink use the term "halo." Is there an actual difference between star halo and star diffraction effect (aside from spikes, of course) that we should learn? I've been considering them the same thing really.

[admin]

Thanks again Ivo. That is interesting.

So each channel is really adjusted independently. But I presume it is the 3D nature of each pixel that makes it appear that the effect is relativistic. Well, it guess it actually truly is relativistic the way the three channels create final pixel colors.

And the fact the multiplier is applied at a linear data level, which we can assume is a pretty low absolute amount, is why one can have multipliers like 20.0 at the max. Though both of those facts would make me start wondering about quantization. Must think more.

Indeed, it's all purely to change the strength of a channel relative to the others.
For the purpose of bog standard RGB color balancing, small factors (<3.0) are almost always sufficient (RGB filters roughly attenuate the signal the same amount, ditto for R, G and B response of the underlying sensor). Terrestrial imaging assumes "white" light being absorbed in some parts of the spectrum more than others, and the remainder reflected.

Objects in space are a whole different ball game, most of the time they don't reflect but emit light themselves (often at very specific wavelengths). Add to that picking and choosing very narrow slivers of the spectrum, and the discrepancies between the channels can be much bigger (infamous example the crazy strong Ha vs most other bands in, for example SHO or HOO composites).

Higher multiplication (or attenuation) factors are often needed to balance the fainter signal in the other bands vs the overwhelming signal of other band(s).

I played with Synth some but am somewhat lost, so more work to do there. I was able to put in all the correct parameters for aperture, focal length, and on page two the proper field of view, but it still seemed pretty mask dependent and while I could get a pretty close match, it wasn't spot on. For some stars the synth spikes would be much greater than reality, for others the opposite. And as far as the star itself, well, I'm a bit uncertain there. Might need to play more with brightness, gamma, and so on. Also not sure where I can find what it is showing me about the PSF that might be useful for how to set the Airy disk in SS.

I wouldn't spend too much time with the Synth module for the purpose of processing an image. It is mostly meant as an educational tool to show you what goes into a PSF as we ultimately see it in out images. Most of the compositing is decidedly less scientific/correct/useful, except choosing the blur (very close to how the atmosphere works). The one thing that is very important (and pertinent to the SS module / Airy Disc size setting) there though is this;

synth2.jpg (83.69 KiB) Viewed 4034 times

(incorrect size setting given the image - this setting is more suitable for a narrow field image)

synth1.jpg (82.88 KiB) Viewed 4034 times

(more plausible setting, suitable for a widefield)

Notice how the stars look very different, purely because of the scale (size) of the - otherwise identical - PSF that is applied. The takeaway should be that in the SS module the size of the Airy Disc (a specific/universal kind of PSF - that of a circular opening) is just as important. At the very least, you should try to match the Aity Disc size to the angular size of the image somewhat.

The Synth module just lets you go nuts with PSF modelling (and shows you what actually goes into the mix of how a PSF comes about), the SS module not at all (you get a single parameter), but the principle of matching PSF scale to image scale remains the same.

The SS module needs a plausible PSF size for the image scale for completely different reasons, and doesn't need a perfect match for your optics, but at least a ballpark scale/size will help it.

Does that make sense?

I maybe use the term halo all the time myself. Should it be limited instead to just artifacts of filters and the like? But if it is just the "normal" halo, even if using the term halo, I still have come around to understanding them as part and parcel of the optics and what was legitimately captured. And is one reason I raise my eyebrow at too much halo-fixing or goodness, star removal. The halo pixels were likely burned out somewhat by the stars, it's set into the data that way, so not like one can remove the halo and now see what was behind it.

However, even things like the ST module description for Shrink use the term "halo." Is there an actual difference between star halo and star diffraction effect (aside from spikes, of course) that we should learn? I've been considering them the same thing really.

A true halo is typically caused by something very different to what astronomers deal with. However the sort of signature reminiscent of a halo (e.g. clear boundary or ring), of course, definitely happens in astronomy. For instance, a pure Airy disc pattern does have discernable rings, and depending on the optics, the first two or three rings may be indeed be visible if a star is sufficiently bright.

Another halo-like phenomenon can be caused by secondary reflections, essentially adding a small ghost image of the circular opening on top of a bright star. Where there is a central obstruction (e.g. for a secondary mirror), this can really start giving the impression of a ring (for examples, see here, here and here).

So, if you want to be pedantic, diffraction patterns are definitely not halos, though in a handful of cases, some diffraction patterns may be reminiscent of halos.

[Mike in Rancho]

Thanks for the detailed info, Ivo.

All starting to make sense now on the bias sliders. And yes I think only one time - not my own data - did I have to max a slider out yet still thought there was more there if I could keep sliding. Indeed a filter scenario. Might have been trying to reveal the squid in the bat, without masking it into existence. I tried to make up for it some by re-weighting the composite module, but it was a lost cause regardless.

I have not gone back to Synth. It was fun for a night, and just as you say, when I looked up my image scale and plugged in the right size, things were much improved in matching the underlying image. So, if I am reading you correctly, that is akin to the Airy Disk percentage we use all the time in SS. Are there any ballparks, and is the slider linear? Just kind of wondering what FOV the 50% default is contemplating. And far left, low percentage, would perhaps be something like a wide angle lens summer Milky Way shot, while far right, high percentage, a long focal length or deep drop? For example Russ just posted up an image that is about 0.65 degree tall, which might be the kind of thing to warrant...90%?

Thanks for the halo links. Those Newtonian effects are brutal, yuck! Glad I haven't made any of those...yet. So true halo is indeed an outer ring and not what we generally mean. Nor can I imagine that I have ever resolved and imaged any true Airy disk rings, though I think I can see them in the iconic Pillars of Creation. So really what we are talking about is just diffraction, whether refractor or reflector, spreading the starlight out, probably keyed to brightness level, and preferably circular and on-center. Newts get the added bonus of diffraction spikes on top of that. Am I getting closer?

That said I haven't considered them much of a bother. The right AutoDev keeps them manageable. And it's where a lot of color is.

[admin]

So, if I am reading you correctly, that is akin to the Airy Disk percentage we use all the time in SS.

Pretty much.

Are there any ballparks, and is the slider linear?

It's indeed linear.

Just kind of wondering what FOV the 50% default is contemplating. And far left, low percentage, would perhaps be something like a wide angle lens summer Milky Way shot, while far right, high percentage, a long focal length or deep drop?

That's it!

For example Russ just posted up an image that is about 0.65 degree tall, which might be the kind of thing to warrant...90%?

Indeed. Of course Focal length and aperture factor in as well, in order to establish something that is "about right".

So really what we are talking about is just diffraction, whether refractor or reflector, spreading the starlight out, probably keyed to brightness level, and preferably circular and on-center. Newts get the added bonus of diffraction spikes on top of that. Am I getting closer?
[/qote]
Spot on!