Color: Physical or Physiological?

[Lotz]

How Narrowband emitters really do look like...

This image shows a comparison of conventional color rendering (left side) and the physically correct color rendering (right side). If you have ever observed M42 in a larger aperture instrument under good skies, you will remeber the grey-blue greenish tint of the outer regions of the Orion nebula, as you can see them in this image.

Monochromatic stimuli like Ha or OIII can not be adjusted by photometric color calibration methods, because those methods work based on white point calibration and therefore are only applicable for continuum emitters (i.e. desaturated colors) like stars, star clusters, galaxies or reflection nebula. Emission nebula, however, have to be numerically assigned the correct color, because they will be found on the very border of the color space and furhtermore are even outside the color gamut of dipslays and printing devices.

I developped this methodology over the past several years, beginning with a problem of colorimetric matching of different display technologies in one vehicle dashboard. Later, I extended this for narrowband emitters which can ,in a next step, perfectly be integrated into the astronomical imaging process. The last weeks being very cloudy gave me the time, to compile all my calculations and scripts into one paper.

To find out more and to learn about the necessary color transformations, you may read my paper on that topic, where I brought all the underlying mathematics together:

http://jupiter.n-t-l.de/DownloadReposit ... ations.pdf

If you don't want to bother too much with theory, this does not matter: At the end of the paper you will find a table with the final results for all common narrowband emitters relevant in astrophotography! There you have the recipe, which color to assign to your narrowband channels. It is not up to one's taste, which color e.g. the Rosette nebula should get. It has its clearly defined color, which can be assigned to the Ha channel, in this special case.

CS

Markus

[Chris Peterson]

How Narrowband emitters really do look like...

This image shows a comparison of conventional color rendering (left side) and the physically correct color rendering (right side). If you have ever observed M42 in a larger aperture instrument under good skies, you will remeber the grey-blue greenish tint of the outer regions of the Orion nebula, as you can see them in this image.

Monochromatic stimuli like Ha or OIII can not be adjusted by photometric color calibration methods, because those methods work based on white point calibration and therefore are only applicable for continuum emitters (i.e. desaturated colors) like stars, star clusters, galaxies or reflection nebula. Emission nebula, however, have to be numerically assigned the correct color, because they will be found on the very border of the color space and furhtermore are even outside the color gamut of dipslays and printing devices.

I developped this methodology over the past several years, beginning with a problem of colorimetric matching of different display technologies in one vehicle dashboard. Later, I extended this for narrowband emitters which can ,in a next step, perfectly be integrated into the astronomical imaging process. The last weeks being very cloudy gave me the time, to compile all my calculations and scripts into one paper.

To find out more and to learn about the necessary color transformations, you may read my paper on that topic, where I brought all the underlying mathematics together:

http://jupiter.n-t-l.de/DownloadReposit ... ations.pdf

If you don't want to bother too much with theory, this does not matter: At the end of the paper you will find a table with the final results for all common narrowband emitters relevant in astrophotography! There you have the recipe, which color to assign to your narrowband channels. It is not up to one's taste, which color e.g. the Rosette nebula should get. It has its clearly defined color, which can be assigned to the Ha channel, in this special case.

M42_comparison.jpg

CS

Markus

The problem is that "color" is physiological, not physical. There is no "correct" mapping for that which we cannot see.

[Lotz]

The problem is that "color" is physiological, not physical. There is no "correct" mapping for that which we cannot see.

Hello Chris,

there I have to strongly disagree. For sure, perception is physiological, but color is NOT! Even, there is a clear relation between the primary stimuli and the color, which shall be shown. That is, by the way, the basis of all color calibration, as photographers do it with their monitors. for example.
Color IS measurable, colour CAN be described mathematically, and color CAN be synthesized reproducably. How an individuum perceives it, is a completely different story. Even a person seing a certain color completely different, would also see the generated color completely different, but still the same as the original real color would have been visible to him.

CS

Markus

[Chris Peterson]

The problem is that "color" is physiological, not physical. There is no "correct" mapping for that which we cannot see.

Hello Chris,

there I have to strongly disagree. For sure, perception is physiological, but color is NOT! Even, there is a clear relation between the primary stimuli and the color, which shall be shown. That is, by the way, the basis of all color calibration, as photographers do it with their monitors. for example.
Color IS measurable, colour CAN be described mathematically, and color CAN be synthesized reproducably. How an individuum perceives it, is a completely different story. Even a person seing a certain color completely different, would also see the generated color completely different, but still the same as the original real color would have been visible to him.

CS

Markus

Color is absolutely physiological. For example, if you take a single pure wavelength, we will observe it as a wide range of colors depending on intensity. I worked in a color perception lab for a couple of years. It's very complex, and we really can't clearly define it mathematically.

[Lotz]

Hi Chris,

it is not color, it is perception, which is physiological! Thats why you worked in a color PERCEPTION lab. I, however, am still working in a color-measurement and color/emitter calibration lab. You will never observe e.g. a monochromatic yellow as a different color or variety of colors, unless, how bright the stimulus will be. For sure, if you are entering mesopic or scotopic vision, then color perception totally drops out. On the other hand, if you overstimulate the eye, then you will see odd color shades, as when directly looking into the sun. But let's assume we are in the steady range of the eye as a receptor.

[Chris Peterson]

Hi Chris,

it is not color, it is perception, which is physiological! Thats why you worked in a color PERCEPTION lab. I, however, am still working in a color-measurement and color/emitter calibration lab. You will never observe e.g. a monochromatic yellow as a different color or variety of colors, unless, how bright the stimulus will be. For sure, if you are entering mesopic or scotopic vision, then color perception totally drops out. On the other hand, if you overstimulate the eye, then you will see odd color shades, as when directly looking into the sun. But let's assume we are in the steady range of the eye as a receptor.

No, it is color that is perceptual. Color is defined as a physiological property. Your work is interesting, but I disagree that it is possible to define any sort of "correct" color for narrowband sources displayed in any image. All you can do is define a workflow or algorithm that is consistent.

[Lotz]

And this consistency is defintely beneficial, as it gives us an instrument to avoid NGC7000s having all colors from orange to purple, just to the individual photogapher's taste.
But believe me, if you display my Ha color on a monitor, it is measurable, that this specific color does have the least delta E to a 656nm emitter of any color displayable on that display. So, not being perefect (I've shown why that's not possible), it is the best possible representation of that color and herewith the nearest possible approximation.

CS

Markus

[Chris Peterson]

And this consistency is defintely beneficial, as it gives us an instrument to avoid NGC7000s having all colors from orange to purple, just to the individual photogapher's taste.
But believe me, if you display my Ha color on a monitor, it is measurable, that this specific color does have the least delta E to a 656nm emitter of any color displayable on that display. So, not being perefect (I've shown why that's not possible), it is the best possible representation of that color and herewith the nearest possible approximation.

CS

Markus

Ha is only some percentage of red on a monitor. It's perceptual color depends on brightness. And it is indistinguishable from any nearby narrowband lines.

[Lotz]

As my intention is not to have any misleading discussions here, I would highly appreciate that one of the moderator team would simply place my post back to the submission thread, because my main intention was to submit this image and its explanation for APOD.

By the way, Ha - for example - isn't even a small percentage of a monitor's red as a monitor's red is defined by typically 620nm with a FWHM value of 5nm, so it is simply not in the spectral distribution of a monitor's red.

CS

Markus

[Chris Peterson]

As my intention is not to have any misleading discussions here, I would highly appreciate that one of the moderator team would simply place my post back to the submission thread, because my main intention was to submit this image and its explanation for APOD.

By the way, Ha - for example - isn't even a small percentage of a monitor's red as a monitor's red is defined by typically 620nm with a FWHM value of 5nm, so it is simply not in the spectral distribution of a monitor's red.

CS

Markus

You misunderstand me. The wavelength of Ha contains no green or blue signal from typical display primaries. It can only be represented by red. Imperfectly, of course. There are emission lines from different species that we can see as different with our eyes (say, with a gas tube) that cannot be rendered as different on a monitor.

[Lotz]

You misunderstand me. The wavelength of Ha contains no green or blue signal from typical display primaries. It can only be represented by red. Imperfectly, of course. There are emission lines from different species that we can see as different with our eyes (say, with a gas tube) that cannot be rendered as different on a monitor.

Unfortunately, I have again to dissent here. You missed one important point of color synthesis in display systems. If we'd say, pure red is the best, how you could represent Ha on a monitor, than, on the other side of the spectrum, pure blue would be the best way to represent violet. Violet, however, can be very good approximated by blue with some addition of red. On the other side of the spectrum, to approximate red longer than the primary stimulus is, you have to add some blue.
Thats the concept of the color triangle and the basis of additive color mixing. For sure, physically, you get a red emission peak and a blue emission peak, but based on the ratio of those peaks you get something looking like longer red or shorter blue (i.e. violet).

Coming to your gas tube example: I totally agree, there will be wavelength pairs, which you can not render differently on a screen, but there will be also wavelengths pairs, which you can clearly render differently on a monitor, That only depends on where in CIE1931 those pure colors are located, how far they are separated and how large the color depth of your imaging system is. But selling OIII somewhere between violet blue and acid green, depending on who made the image, is somewhat odd.

I am quite sure, this is, what you mean, when saying color is physiological. It looks e.g. violet, but indeed is just a blue and a red emisison peak. The same is for white: it can be synthesized by 3 very narrow peaks, but could as well be a broadband continuum. But that's how our visual apparatus works and has nothing to do with color itself. Coming to Ha or OIII ( whose hue is at least not outside the color triangle, and therefore maybe more intuitive): There is exactly ONE (R;G;B) triple (with R being zero in the case of OIII), that generates a color hue, which is a best fit to the color hue of a pure OIII emitter. It's for sure not perfect - I clearly state in my paper, that narrowband emitters CAN NOT be perfectly represented on any kind of currently available imaging devices - but it is the best approximation for those colors and there is an explicit color tupel for each wavelength and each color gamut. Yes, it is not perfect, as I stated in my paper, but those colors (a) give consistentcy and (b) when displayed on a monitor have the least deviation from the real emitter.

CS

Markus