Measuring dominant wavelength?

[METsen]

For a personal study I would like to measure in photos the dominant wavelength of the sun image near its setting. Do you know of a program which can show this value e.g. for a selected pixel region?

 

[photoguy99]

If you want the actual wavelength, it cannot be done with a consumer camera which is not measuring wavelengths at all but rather approximating the colors as perceived by the human eye.

If you just want the color, I don't know of a program that will do it with one click, but you could use any reasonable image editor. Cut the area out and make a new image if just that. Apply blur until the color is quite uniform. Use a color picker tool in the editor to pick that color and examine it.

 

[METsen]

If you want the actual wavelength, it cannot be done with a consumer camera which is not measuring wavelengths at all but rather approximating the colors as perceived by the human eye.

Yes, but this approximated color corresponds to some spectral distribution and its corresponding dominant waevelength. I really need the dominant wavelength of the spectral distribution of a selected pixel area or of the matching color.

 

[photoguy99]

A perceived color can correspond to infinitely many different actual spectra. Too much spectral information is lost, you cannot back track to a 'dominant wavelength' from the RGB data a camera captures.

Key phrases for searches:

Color vision
Perceptual color
Physical color

 

[METsen]

... you cannot back track to a 'dominant wavelength' from the RGB data a camera captures.

Yes, the original spectrum is indeed lost. I thought however that in a similar way as the perceived color is reconstructed from the RGB data, it would also be possible to estimate the original dominant wavelength from them. (BTW, 'dominant wavelength' is a defined and calculated value within a spectral range.)

 

[photoguy99]

Nope!

Not in general, anyways.

If you happen to know that the dominant wavelength has to be one of a small family of candidates, you might be able to figure out which one from RGB data. Maybe. It will depend, I'm pretty sure, on the details, if it's possible at all.

 

[METsen]

Thanks, photoguy99, for all your explanations. Too bad that it is not possible, it would have given me an alternative means.
Regards, METsen

 

[TCampbell]

You actually can capture the spectra of the sun with a regular camera, but it helps to have an unfiltered camera (preferably a monochrome camera). You need a diffraction grating to do it ... and a bit of software: RSpec / Real-time Spectroscopy

But you also need to get the sun reduced to a tiny slit (it cannot appear as a "disk" or it won't work.)

This works because the diffraction grating is a kind of prism that smears the star into it's rainbow spectra. You get a single "point" representing the star (or the slit representing the sun) and the "smear" the side is the spectra. The computer software analyzes the "smear" because it turns out bits of it will be brighter or dimmer (representing the amount of energy at each specific wavelength). The distance of any given point of the "smear" from the point source (or slit) can be measured in pixels, but once calibrated (usually by finding the Hydrogen b absorption line to calibrate because you get a noticeable dip) it can then convert the pixel distance to angstroms.

At that point you get a graph showing the strength and specific wavelength of the entire spectra and you can do this with any camera but the built-in UV & IR blocking filter taints the accuracy. That's why it's better to have a monochrome camera or a modified color camera.

 

[photoguy99]

Nice! That's very cool.

Can't you also do it with a shoebox at this point?

 

[snowbear]

I usually go with "orange-ish"

 

[METsen]

You actually can capture the spectra of the sun with a regular camera, but it helps to have an unfiltered camera (preferably a monochrome camera). You need a diffraction grating to do it ... and a bit of software: RSpec / Real-time Spectroscopy

Thank you for this tipp! That can indeed become useful one day, but unfortunately not for what I am looking for at the moment, wondering whether it would be possible to extract additional information from the colours in a large collection of existing photos which have not been made with the intention to extract later some spectral information from their colours.

An other place indicated this link (especially this one) showing that some researchers appear to convert RGB to wavelength within certain restrictions. I do not know whether one of their solutions became finally a general applicable tool in a photo editing program; do you know more?

Thanks to all of you
METsen