Camera Sensitivity to Ha and "natural" Colour?

[PRejto (Peter)]

I'm posting this question here because ultimately it involves equipment choices.

I ran into this discussion today and I wonder what others think? It boils down to an argument that astro-cameras unnaturally (compared to our eyes) respond to Ha far too much and this results in overly saturated red images of nebula. We often claim that colour balance in astrophotos is "real" and just boosted so that if our eyes were more sensitive to colour that this is what we would see...similar to the argument that grass doesn't look green at night but if we boost it it will look green.

Personally I think there is validity to this argument. Adding additional Ha in the R channel which happens by virtue of the CCD/CMOS sensitivity does in fact create an unnatural color image if we want to honestly create an image that reflects what our unaided eyes "might" see.

https://clarkvision.com/articles/col...rstellar.dust/

Peter

Edit: Thinking further about this, I wouldn’t want to give up the detail that pops out in Ha. Why not collect red data with a filter that more matches what our eyes might see if they were simply more responsive. That would give a good colour balance. Then blend in Ha to luminance for the detail? I suppose we do that anyway in RGB imaging but perhaps with too much red.

[Ryderscope (Rodney)]

Thanks for raising this issue Peter, it's a great topic to explore. I don't have an exhaustive response as yet but can raise a couple of quick points and questions.

"I'm posting this question here because ultimately it involves equipment choices."

I am wondering what equipment choices would be reliant on a specific approach to colour management?

As for the arguments in the Clarkvision article:

Roger Clark's article is very detailed and is very informative on the science behind understanding colours in space and specifically in the Milky Way as they are applied to different chemical elements and dusty regions. However, I think that there is a gap between an attempt to codify what is "correct" colour management by referring to the science of colour and the reality of how we perceive colours as human beings. Aesthetically when we some colour groupings side by side they either look or they don't. Looking at the widefield Orion image in the article, the colours in this image may be correct according the Mr. Clark's 'natural' colours but you would not see too many versions of this field using the same colour balance.

The problem is that there is no agreed authority and what is correct colour and what isn't in an astrophotograph. I've seen, for example, images commented on in competitions as the colour not being quite right but with a lack of follow up to say what exactly that means and what the colour should have been or should not have been.


Edit: Thinking further about this, I wouldn’t want to give up the detail that pops out in Ha. Why not collect red data with a filter that more matches what our eyes might see if they were simply more responsive. That would give a good colour balance. Then blend in Ha to luminance for the detail? I suppose we do that anyway in RGB imaging but perhaps with too much red.

Exactly, we are making the image do what we want it to do. We want to see lots of detail of what is there and a great way to do this is to add colour to it. For what it is worth, I have recently started to do what you describe in mixing the Ha with the Luminance to use as the Luminance channel so that it carries more detail.

And finally, the Clarkvision article discusses the emission lines in different ionised Hydrogen and the scattering of light by minute dust particles in the Milky Way. There is, however, no mention of star colours which is another rabbit hole that we could go down. Whist there is a well defined system for classifying stars based on the colour (temperature), there remains a lot of subjective control that can be applied in post processing when it comes to enhancing the star colours.

Clear skies,
Rodney

[PRejto (Peter)]

Hi Rodney,

Thanks for your detailed and thoughtful reply! You make a number of great observations.

Well, yes, colour is pretty subjective. We cannot even agree that humans "see" the same thing at all from person to person, but we assume that we do and I think it is a reasonable leap as a starting point.

Here is my attempt to put this in perspective. Suppose I told you that I have a great new blue filter that captures amazing UV and that your images will now look so rich in blue. Would you accept that? Probably not, so I see no difference with the red filter. Our eyes don't have great sensitivity to Ha but the cameras sure do and the filters just transmit it all...and it overwhelms the red signal. So, why not use a filter (equipment!) that transmits the proportion of light that our eyes naturally respond to? Every normal digital camera sold for everyday use does this, but we insist on removing that filter so these cameras can see more Ha...and the white balance goes kaput. Images with this everyday white balance would not be devoid of red...in fact, t seems that other colours and shading would emerge,,,oranges, browns, etc that currently are all just red.

Now, lets not confuse stretching and detail with colour. There is much that is artificial and manufactured in our images. My question only concerns whether we as astrophotographers are deluding ourselves into thinking that RGB imaging is presenting "true" colours. i.e., "if only our eyes were more sensitive to the coloured light, this is what we will would if the light is boosted enough to perceive it." To beat a dead horse, if I took a picture of grass at night with an astro-camera and processed it normally and then said, this is what grass looks like if only there was more light we would laugh. But, we do exactly that every day with objects in the sky. This is why I want to experiment more with a red filter that more matches the human eye's response. I have no problem collecting Ha and adding to luminance and creating a totally artificial picture. Or, maybe I should just give up this quest all together and say it's all artificial and in the end "art" and be done with it.

[Startrek (Martin)]

Peter,
Thanks for posting
Personally I don’t subscribe to what Roger Clarke or even Tony Hallas say in regard to Astrophotography

One person I listen to what Ivo Jager says, an extremely clever and factual individual who developed Startools

Refer to link below….

https://forum.startools.org/viewtopic.php?t=912

Colour in Astrophotography is very subjective and always will be. Nothing wrong with Ha dominant images ,to me it’s the detail and resolution that takes priority over colour. Although colour especially Narrowband does provide an image with the Wow factor.
If I looked in my eye piece at M42 and saw a beautiful coloured nebula with shades of reds , greens and blues then that could be the basis of natural colour in celestial objects. Unfortunately my eyes just see a black and white picture show ( which is still beautiful)

Cheers
Martin

[Merlin66 (Ken)]

Peter, et al,
As you probably know I only do spectroscopy after many years of visual and astrophotography.
There's no such thing a true colour in the images we record.


The spectrum of the various objects at least comes somewhere towards a detailed distribution of the light received. There are issues of interfering dust clouds, Doppler shift etc.
I can correct for the camera response and sometimes, if data is available, the dust absorption. I could, based on the scientific data I have on the response of the normal human eye ("Researches in colour vision", Sir W Abney.) make the necessary corrections across the spectum to give a "visual colour representation"

I honestly don't think the effort would be worth it.
Ken

[AdamJL]

There's also the issue of the inverse-square-law. Most of the images we create don't match we'd see if we were there. Actually, what does "there" even mean? No software, as far I can tell, reverses the inverse square law to create an image based on brightness (and therefore colour) that would be true to a given distance from the object.

Plus just adding more exposure time means you're bringing out fainter details and colours in an object that we might not see at all with our eyes because integration boosts them.

[bojan]

Irrelevant, in my opinion.
AP is kind of technical photography (IMHO), so I am interested only in details, revealing the physical nature of the object, and not the visual appearance or "artistic" component..

That means: high resolution and lots of details.. achievable by larger aperture and better visibility (lowest possible LP, highest possible altitude), and appropriate level of post-processing (so no "details" added that are not there in reality).

Everything else or more than above is fake in principle...

[AdamJL]

Originally Posted by bojan

Everything else or more than above is fake in principle...

Everything is fake as soon as you stretch your data. That's for both colour and luminance.
The degree of what we are comfortable with when it comes to that fakery is different to everyone, but it's all still fake.

I also agree with Ken.. there's no such thing as "true" colour that we can all agree on in this hobby. Even if you are capturing Ha data, the bandpass is too wide and it does bring in more red than would be if you were able to solely isolate true Ha signal. And then that entire spectrum becomes "Ha", even though only a small part of it is.
Filters are getting better at filtering out signal that doesn't match the target wavelength, but it still gets mapped as Ha. Does that then boost the red in an image artificially considering you'd still need someway to capture RGB (through broadband filters or CFAs).. sure it does!

[bojan]

Originally Posted by AdamJL

Everything is fake as soon as you stretch your data ....

Not necessarily.. you can still do accurate enough photometry on images with stretched histogram.

However, using AI (even just deconvolution) to bring out details actually destroys data, and that is fake.

But some can consider that as "art"..

[Bassnut (Fred)]

i don't really understand the problem. If you think Ha is too strong (or "unnatural") in the red signal, then capture RGB and 3nm Ha separately. Subtract (say in PS) Ha from the red signal and then blend in the separate Ha channel to taste (if at all). I must say though, a nebula pic without Ha at all would be boring AF regardless of how much it conforms to an "a realistic human eye response experience"

[PRejto (Peter)]

Wow, I appreciate all the responses that suddenly came in today!

The take away that I'm getting is that colour is entirely subjective and there is no such thing as "real." I don't really disagree, but when I accidentally read the article by Clark it made me stop and think and I spent a few days mulling all what he said. And, why not?

I don't agree with some of the comments that the photos he posted (processed to tame the reds) are "boring." They are different than the usual fare, and the reds show an interesting distribution of colours usually rendered as all red. Anyway, it's probably not worth beating this dead horse anymore!

Maybe anyone who comments on any posted photos in the deep sky section should refrain from making any comments about colour. To the contrary, quite a few posts pick on colour if it doesn't happen to fit ones own notion of what looks "real!" And, it's pretty hard to deny that Australia's most famous astro imaging competition puts quite an emphasis on colour that is supposedly "real and natural." So, while there may be agreement on one hand that colour is fake, it's also pretty clear that somewhere deep in the collective psyche their exists a notion that colour needs to look a certain way or it isn't right. We just don't all agree one what that means.

Peter

[Peter Ward]

I hesitated on chiming in here, but would suggest instead of just reading Clarke's web page, that you look at the work of Helmholtz and Maxwell who established colour theory about 175 years ago.

Malin's decidedly more contemporary "Colours of stars" is another excellent reference.

To suggest a "standard" off the shelf camera renders accurate nebula colour is nonsense. The white balance setting on a camera is there for a good reason. Ignoring the colour temperature of the source can easily bias the scene from red to blue, which if not compensated for, renders the scene a very "un-natural" fashion.

DSLR's are also fitted with cutoff filters that intentionally attenuate the IR plus red end of the spectrum and do not pass a significant portion of any h-alpha flux to their sensors, which would be swamped by even longer wavelengths. Balanced colour without this blocking filter is not possible. Owners of "full spectrum" modded cameras (self included) can attest to this.

So how does this relate to "natural" colour?

Normal human vision goes from photopic (daylight) mode driven by retinal cone cells (read colour sensitive) to scotopic (night time ) vision driven by rod cells which are most sensitive to wavelengths of around 498*nm (blue-green) and are insensitive to wavelengths longer than about 640*nm (red-orange).

In short humans don't perceive much colour viewing a landscape illuminated by a moonlit sky and as for scotopic viewing of faint h-alpha, well, forget it.

But, is the colour information still there? Of course it is. Our eyes can see 654nm wavelength red light ( e.g. observing the Sun through a h-alpha filter ) and deep violet, plus a myriad of hues in-between. Lack of illumination is not a good reason to assume what you perceive in the dark is accurate or "natural" colour.

The spectral flux emitted by nebulae while faint can be measured. RGB filtered images from the likes of the AAO used a matrix of ND filters to determine just how much flux was present in red blue and green exposures. This allowed RG&B exposures to be accurately calibrated and combined to deliver a faithful full colour image. Sure, there may be a slight calibration bias, but for example, the Tarantula Nebula emits mainly red light. Rendering it as blue-green is just plain wrong, yet this is exactly how most off the shelf cameras render it. By switching to an extended red spectrum camera, such as a Canon EOSRa, the nebula is recorded correctly as red. Interestingly the same camera when pointed at a daylight illuminated scene records "normal" colours with remarkable fidelity, albeit not quite as well as a standard model with daylight illumination.

If you really want to strive for accuracy, then a colour target chart can be used to calibrate your imaging system by varying RG&B exposures of the target, that when combined, best match the original chart. Once these exposure ratios are known, you can image deep sky objects with reasonable confidence that the results will indeed be accurate. Imaging software such as Pixinsight can also calibrate your colour data by using a star of known colour temperature to correct the image. Howver, care should be taken not to use field stars that are saturated (i.e. burnt out) with this method.

All that said, narrow band data can also be used to either enhance RGB images or be combined to produce SII, Ha and OIII imagery. But while adding H-alpha (i.e. red) to green or blue channels might produce an interesting result, it ignores the physical reality. A bad move IMHO. Just because you can, doesn't mean you should.

With narrow band imaging it's dealer's choice as there is no "natural" colour filter combination that can give you a full-spectrum image. However I'd suggest using an eye-pleasing palette rather than a combination that results in muddy brownish result.