Measuring 'Dominant Wavelength'?

[Marcel Tschudin]

Hello everyone,

I am new here. I am wondering whether I could use ImageJ (or an other program) for measuring in photos the 'Dominant Wavelength' of the colors within a selected pixel area. I provide here some further explanations because I am not sure whether what I intend to do would actually even be possible with photos.

I would like to estimate the sun's 'Dominant Wavelength' in photos of the setting sun. For a detector like the human eye the 'Dominant Wavelength' would result from the sun's spectrum after passing the atmosphere and after passing the eye's spectral detector efficiency. It would be calculated from the detector's spectrum as Ldom, with the radiation intensity, I, at a certain wavelength, L, in increments, dL, over the visible spectrum as a ratio of two sums (integrals):
Ldom = Sum(I*L*dL) / Sum(I*dL)
(Because 'Dominant Wavelength' could be misinterpreted others suggest to call this the 'Balanced Wavelength' instead.)

Consumer cameras do not record the spectrum, they rather approximate the detected spectral content, i.e. the color perceived by the human eye, with the RGB information. Would it now be possible to estimate the original 'Dominant Wavelength' from the available RGB information in the photo? If yes, do you know if ImageJ (or an other program) provides such a feature or a similar one?

Thanks,
Marcel

--
ImageJ mailing list: http://imagej.nih.gov/ij/list.html

[Herbie-4]

Marcel,

for wavelengths you would need an additional dimension lambda, i.e. a
spectrometer at every pixel.

At best RGB-images give you three values of this lambda-dimension.
Furthermore, you can't be sure how these values are generated. They
depend on the properties of the three color filters of the camera that
usually spectrally overlap. Finally, most digital color cameras don't
use three sensors (RGB) but a Bayer color-mask which necessitates that
the RGB values at every pixel are at least partially computed from the
surrounding pixels.

That said, there is little to no chance that you get the desired
spectral information.

HTH

Herbie

::::::::::::::::::::::::::::::::::::::::
On 27.01.15 20:35, Marcel Tschudin wrote:

> Hello everyone,
>
> I am new here. I am wondering whether I could use ImageJ (or an other
> program) for measuring in photos the 'Dominant Wavelength' of the
> colors within a selected pixel area. I provide here some further
> explanations because I am not sure whether what I intend to do would
> actually even be possible with photos.
>
> I would like to estimate the sun's 'Dominant Wavelength' in photos of
> the setting sun. For a detector like the human eye the 'Dominant
> Wavelength' would result from the sun's spectrum after passing the
> atmosphere and after passing the eye's spectral detector efficiency.
> It would be calculated from the detector's spectrum as Ldom, with the
> radiation intensity, I, at a certain wavelength, L, in increments,
> dL, over the visible spectrum as a ratio of two sums (integrals):
> Ldom = Sum(I*L*dL) / Sum(I*dL) (Because 'Dominant Wavelength' could
> be misinterpreted others suggest to call this the 'Balanced
> Wavelength' instead.)
>
> Consumer cameras do not record the spectrum, they rather approximate
> the detected spectral content, i.e. the color perceived by the human
> eye, with the RGB information. Would it now be possible to estimate
> the original 'Dominant Wavelength' from the available RGB information
> in the photo? If yes, do you know if ImageJ (or an other program)
> provides such a feature or a similar one?
>
> Thanks, Marcel
>
> -- ImageJ mailing list: http://imagej.nih.gov/ij/list.html
>

--
ImageJ mailing list: http://imagej.nih.gov/ij/list.html

[Marcel Tschudin]

Thank you, Herbie, for your explanations. It's too bad that I cannot follow further this option. I would have thought that somehow the reverse should be possible of what one does with graphical representations where the values of a parameter are shown in different colors with the color spectrum indicating the value range.
Marcel

--
ImageJ mailing list: http://imagej.nih.gov/ij/list.html

[Chuck Harrison-2]

My thoughts:

1. As Herbie writes, this task is impossible without further constraints.
The camera doesn't give you enough information.

2. I think the best you could hope for would be to do a characterization of
a single camera in this specific task. This would require a series of
sunset measurements using an independent instrument (spectroradiometer)
which DOES measure what you want, and simultaneous photographs. Then you
could develop an empirical correlation.

3. Do look for a different term; "dominant wavelength" has a precise
meaning in colorimetry. "Power averaged wavelength" might be a suitable
term for your computation.

4. Your computed value makes no reference to the human eye response, so I
hope you are not expecting it to have a consistent relationship to what the
setting sun looks like when observed by a human being.

Cheers,
  Chuck


On Tue, Jan 27, 2015 at 11:35 AM, Marcel Tschudin <
[hidden email]> wrote:

> Hello everyone,
>
> I am new here. I am wondering whether I could use ImageJ (or an other
> program) for measuring in photos the 'Dominant Wavelength' of the colors
> within a selected pixel area. I provide here some further explanations
> because I am not sure whether what I intend to do would actually even be
> possible with photos.
>
> I would like to estimate the sun's 'Dominant Wavelength' in photos of the
> setting sun. For a detector like the human eye the 'Dominant Wavelength'
> would result from the sun's spectrum after passing the atmosphere and after
> passing the eye's spectral detector efficiency. It would be calculated from
> the detector's spectrum as Ldom, with the radiation intensity, I, at a
> certain wavelength, L, in increments, dL, over the visible spectrum as a
> ratio of two sums (integrals):
> Ldom = Sum(I*L*dL) / Sum(I*dL)
> (Because 'Dominant Wavelength' could be misinterpreted others suggest to
> call this the 'Balanced Wavelength' instead.)
>
> Consumer cameras do not record the spectrum, they rather approximate the
> detected spectral content, i.e. the color perceived by the human eye, with
> the RGB information. Would it now be possible to estimate the original
> 'Dominant Wavelength' from the available RGB information in the photo? If
> yes, do you know if ImageJ (or an other program) provides such a feature or
> a similar one?
>
> Thanks,
> Marcel
>
> --
> ImageJ mailing list: http://imagej.nih.gov/ij/list.html
>

--
ImageJ mailing list: http://imagej.nih.gov/ij/list.html

[Jean-Louis Oneto]

I agree with the others answers you got, that in general there is no
unique conversion possible between RGB and wavelength, nevertheless the
problem can get an approximate solution.
See for examples:
http://stackoverflow.com/questions/5817474/how-to-get-the-wavelength-of-a-pixel-using-rgb
https://www.physicsforums.com/threads/colour-rgb-hsv-hsl-to-wavelength.419793/
http://www.researchgate.net/post/What_is_a_good_way_to_convert_a_RGB_pixel_to_a_wavelength
It would be better in any case to begin with a color calibration of your
camera (by taking pictures of calibrated targets and analyzing the
pictures obtained with a software able to generate an ICC profile (there
are several, both  commercial and freeware).
Hope that helps,
Best regards,
Jean-Louis

On 27/01/2015 21:42, Marcel Tschudin wrote:
> Thank you, Herbie, for your explanations. It's too bad that I cannot follow further this option. I would have thought that somehow the reverse should be possible of what one does with graphical representations where the values of a parameter are shown in different colors with the color spectrum indicating the value range.
> Marcel
>
> --
> ImageJ mailing list: http://imagej.nih.gov/ij/list.html
>

--
ImageJ mailing list: http://imagej.nih.gov/ij/list.html

[Marcel Tschudin]

On Tue, Jan 27, 2015 at 11:44 PM, Chuck Harrison <[hidden email]> wrote:

> ...
> 3. Do look for a different term; "dominant wavelength" has a precise
> meaning in colorimetry. "Power averaged wavelength" might be a suitable
> term for your computation.
>

The original intention was actually indeed to select the balance point
within the spectrum. I would therefore also favour "Balance Wavelength" as
an alternative.


> 4. Your computed value makes no reference to the human eye response, so I
> hope you are not expecting it to have a consistent relationship to what the
> setting sun looks like when observed by a human being.
>

It looks like I have to provide some further details: The detector, eye or
camera, receives a response spectrum from the setting sun (yes, also
depending on the detector's response). Which wavelength within this
spectrum was responsible for the amount of refraction that this light
package actually did hit the detector and not a place above or below it?
One way of estimating it is by simulation, using an appropriate atmospheric
model. I just was wondering now whether the photos themselves, made with a
normal consumer camera, would allow estimating this wavelength. I learned
so far that this does not appear to exist, and that it would likely not
even be possible.

Regards,
Marcel

>
> Cheers,
>   Chuck
>
>
> On Tue, Jan 27, 2015 at 11:35 AM, Marcel Tschudin <
> [hidden email]> wrote:
>
> > Hello everyone,
> >
> > I am new here. I am wondering whether I could use ImageJ (or an other
> > program) for measuring in photos the 'Dominant Wavelength' of the colors
> > within a selected pixel area. I provide here some further explanations
> > because I am not sure whether what I intend to do would actually even be
> > possible with photos.
> >
> > I would like to estimate the sun's 'Dominant Wavelength' in photos of the
> > setting sun. For a detector like the human eye the 'Dominant Wavelength'
> > would result from the sun's spectrum after passing the atmosphere and
> after
> > passing the eye's spectral detector efficiency. It would be calculated
> from
> > the detector's spectrum as Ldom, with the radiation intensity, I, at a
> > certain wavelength, L, in increments, dL, over the visible spectrum as a
> > ratio of two sums (integrals):
> > Ldom = Sum(I*L*dL) / Sum(I*dL)
> > (Because 'Dominant Wavelength' could be misinterpreted others suggest to
> > call this the 'Balanced Wavelength' instead.)
> >
> > Consumer cameras do not record the spectrum, they rather approximate the
> > detected spectral content, i.e. the color perceived by the human eye,
> with
> > the RGB information. Would it now be possible to estimate the original
> > 'Dominant Wavelength' from the available RGB information in the photo? If
> > yes, do you know if ImageJ (or an other program) provides such a feature
> or
> > a similar one?
> >
> > Thanks,
> > Marcel
> >
> > --
> > ImageJ mailing list: http://imagej.nih.gov/ij/list.html
> >
>
> --
> ImageJ mailing list: http://imagej.nih.gov/ij/list.html
>

--
ImageJ mailing list: http://imagej.nih.gov/ij/list.html

[GDC]

Probably the best you can do without a spectrometer would be to convert
the color space of your image to a cylindrical one such as HSV and use
the H (Hue) value to get an approximation of the dominant color (not
wavelength).   Alternatively, you might get better results with a simple
diffraction grating spectrometer in front of the camera ala the public
lab spectrometer.

Greg
On 1/27/2015 2:35 PM, Marcel Tschudin wrote:

> Hello everyone,
>
> I am new here. I am wondering whether I could use ImageJ (or an other program) for measuring in photos the 'Dominant Wavelength' of the colors within a selected pixel area. I provide here some further explanations because I am not sure whether what I intend to do would actually even be possible with photos.
>
> I would like to estimate the sun's 'Dominant Wavelength' in photos of the setting sun. For a detector like the human eye the 'Dominant Wavelength' would result from the sun's spectrum after passing the atmosphere and after passing the eye's spectral detector efficiency. It would be calculated from the detector's spectrum as Ldom, with the radiation intensity, I, at a certain wavelength, L, in increments, dL, over the visible spectrum as a ratio of two sums (integrals):
> Ldom = Sum(I*L*dL) / Sum(I*dL)
> (Because 'Dominant Wavelength' could be misinterpreted others suggest to call this the 'Balanced Wavelength' instead.)
>
> Consumer cameras do not record the spectrum, they rather approximate the detected spectral content, i.e. the color perceived by the human eye, with the RGB information. Would it now be possible to estimate the original 'Dominant Wavelength' from the available RGB information in the photo? If yes, do you know if ImageJ (or an other program) provides such a feature or a similar one?
>
> Thanks,
> Marcel
>
> --
> ImageJ mailing list: http://imagej.nih.gov/ij/list.html

--
ImageJ mailing list: http://imagej.nih.gov/ij/list.html

[Kenneth Sloan-2]

How many assumptions are you willing to make?

The human eye is no more able to compute the “dominant wavelength” of an arbitrary spectrum than a camera is.  But, the human eye did not evolve in an environment of arbitrary spectra.

If you assume “natural light”, or “black body radiation”, then there is some hope.  If you want to do this in the face of arbitrary lighting conditions - then, no - it simply can’t be done.  Either by the human visual system or a traditional camera.  

How many samples are required (across the spectrum)?  Alas, in general the answer is: an infinite number. Any claim that you can use fewer is simply an assertion about the nature of the lighting.

In fact, even the concept of “Dominant Wavelength” involves considerable assumptions.  In fact…in those cases where “Dominant Wavelength” makes sense… then human eyes and conventional RGB cameras can do the job just fine.  The difficulty comes when you try to extend this concept into domains where it doesn’t apply.  “Dominant Wavelength” is a concept that really only makes sense in a color system that is 3D - one where “Hue” is a possible dimension.  It’s really a psychological concept, not a physical one [except to the extent that psychology evolved to match a certain flavor of physical reality].

Confused, yet?  Good!

--
Kenneth Sloan
[hidden email]

> On Jan 27, 2015, at 13:35 , Marcel Tschudin <[hidden email]> wrote:
>
> Hello everyone,
>
> I am new here. I am wondering whether I could use ImageJ (or an other program) for measuring in photos the 'Dominant Wavelength' of the colors within a selected pixel area. I provide here some further explanations because I am not sure whether what I intend to do would actually even be possible with photos.
>
> I would like to estimate the sun's 'Dominant Wavelength' in photos of the setting sun. For a detector like the human eye the 'Dominant Wavelength' would result from the sun's spectrum after passing the atmosphere and after passing the eye's spectral detector efficiency. It would be calculated from the detector's spectrum as Ldom, with the radiation intensity, I, at a certain wavelength, L, in increments, dL, over the visible spectrum as a ratio of two sums (integrals):
> Ldom = Sum(I*L*dL) / Sum(I*dL)
> (Because 'Dominant Wavelength' could be misinterpreted others suggest to call this the 'Balanced Wavelength' instead.)
>
> Consumer cameras do not record the spectrum, they rather approximate the detected spectral content, i.e. the color perceived by the human eye, with the RGB information. Would it now be possible to estimate the original 'Dominant Wavelength' from the available RGB information in the photo? If yes, do you know if ImageJ (or an other program) provides such a feature or a similar one?
>
> Thanks,
> Marcel
>
> --
> ImageJ mailing list: http://imagej.nih.gov/ij/list.html

--
ImageJ mailing list: http://imagej.nih.gov/ij/list.html

[Gabriel Landini]

On Tuesday 27 Jan 2015 19:31:27 Greg <[hidden email]> wrote:
> Probably the best you can do without a spectrometer would be to convert
> the color space of your image to a cylindrical one such as HSV and use
> the H (Hue) value to get an approximation of the dominant color (not
> wavelength).   Alternatively, you might get better results with a simple
> diffraction grating spectrometer in front of the camera ala the public
> lab spectrometer.

Hm... not in HSV space derived from RGB images...
E.g. "Yellow pixels" could be proper "yellow light", or the "sum or red and
green light". You can't reach any conclusions with HSV.

If you want to do this properly, with images, you could use a calibrated
greyscale camera and a bank of narrow band pass filters or one of those fancy
"tunable filters" like the Varispec system.

Cheers

Gabriel

--
ImageJ mailing list: http://imagej.nih.gov/ij/list.html

[Marcel Tschudin]

Thanks to all of you for all your various comments. I only was wondering
whether I could extract additional information from the colours in a large
collection of photos which have not been made for this specific purpose.
The links provided by Jean-Louis show that some researchers appear to
convert RGB to wavelength with certain restrictions. It looks however like
none of their solutions became a general applicable tool in ImageJ (or
another photo program?).

Regards,
Marcel


On Wed, Jan 28, 2015 at 11:37 AM, Gabriel Landini <[hidden email]>
wrote:

> On Tuesday 27 Jan 2015 19:31:27 Greg <[hidden email]> wrote:
> > Probably the best you can do without a spectrometer would be to convert
> > the color space of your image to a cylindrical one such as HSV and use
> > the H (Hue) value to get an approximation of the dominant color (not
> > wavelength).   Alternatively, you might get better results with a simple
> > diffraction grating spectrometer in front of the camera ala the public
> > lab spectrometer.
>
> Hm... not in HSV space derived from RGB images...
> E.g. "Yellow pixels" could be proper "yellow light", or the "sum or red and
> green light". You can't reach any conclusions with HSV.
>
> If you want to do this properly, with images, you could use a calibrated
> greyscale camera and a bank of narrow band pass filters or one of those
> fancy
> "tunable filters" like the Varispec system.
>
> Cheers
>
> Gabriel
>
> --
> ImageJ mailing list: http://imagej.nih.gov/ij/list.html
>

--
ImageJ mailing list: http://imagej.nih.gov/ij/list.html

[Marcel Tschudin]

1) Sorry, I made a big mistake by designating the balance point in the
response spectrum of a detector "dominant wavelength". I was not aware that
this expression is already used in colour science. I noticed also that
others tend to misinterpret it as the statistical modal value. I find
"balance wavelength" is less confusing and relates even better to the
initial purpose, but there might even be better ones.

2) I hoped to obtain from the RGB values of the sun image, which
approximate the perceived colour of it, an estimate for the "balance
wavelength" of the camera's response spectrum. The sun's spectrum outside
the atmosphere is known. With a fairly high probability the result will be
between e.g. 500nm and 700nm. The transmission through the atmosphere can
be estimated using aerosol measurements. But here we start now to estimate
the "balance wavelength" by simulation, and I actually wondered whether I
could obtain from the photos an estimate independent of the simulation.

Marcel

On Wed, Jan 28, 2015 at 6:34 AM, Kenneth Sloan <[hidden email]>
wrote:

> How many assumptions are you willing to make?
>
> The human eye is no more able to compute the “dominant wavelength” of an
> arbitrary spectrum than a camera is.  But, the human eye did not evolve in
> an environment of arbitrary spectra.
>
> If you assume “natural light”, or “black body radiation”, then there is
> some hope.  If you want to do this in the face of arbitrary lighting
> conditions - then, no - it simply can’t be done.  Either by the human
> visual system or a traditional camera.
>
> How many samples are required (across the spectrum)?  Alas, in general the
> answer is: an infinite number. Any claim that you can use fewer is simply
> an assertion about the nature of the lighting.
>
> In fact, even the concept of “Dominant Wavelength” involves considerable
> assumptions.  In fact…in those cases where “Dominant Wavelength” makes
> sense… then human eyes and conventional RGB cameras can do the job just
> fine.  The difficulty comes when you try to extend this concept into
> domains where it doesn’t apply.  “Dominant Wavelength” is a concept that
> really only makes sense in a color system that is 3D - one where “Hue” is a
> possible dimension.  It’s really a psychological concept, not a physical
> one [except to the extent that psychology evolved to match a certain flavor
> of physical reality].
>
> Confused, yet?  Good!
>
> --
> Kenneth Sloan
> [hidden email]
>
>
> > On Jan 27, 2015, at 13:35 , Marcel Tschudin <[hidden email]>
> wrote:
> >
> > Hello everyone,
> >
> > I am new here. I am wondering whether I could use ImageJ (or an other
> program) for measuring in photos the 'Dominant Wavelength' of the colors
> within a selected pixel area. I provide here some further explanations
> because I am not sure whether what I intend to do would actually even be
> possible with photos.
> >
> > I would like to estimate the sun's 'Dominant Wavelength' in photos of
> the setting sun. For a detector like the human eye the 'Dominant
> Wavelength' would result from the sun's spectrum after passing the
> atmosphere and after passing the eye's spectral detector efficiency. It
> would be calculated from the detector's spectrum as Ldom, with the
> radiation intensity, I, at a certain wavelength, L, in increments, dL, over
> the visible spectrum as a ratio of two sums (integrals):
> > Ldom = Sum(I*L*dL) / Sum(I*dL)
> > (Because 'Dominant Wavelength' could be misinterpreted others suggest to
> call this the 'Balanced Wavelength' instead.)
> >
> > Consumer cameras do not record the spectrum, they rather approximate the
> detected spectral content, i.e. the color perceived by the human eye, with
> the RGB information. Would it now be possible to estimate the original
> 'Dominant Wavelength' from the available RGB information in the photo? If
> yes, do you know if ImageJ (or an other program) provides such a feature or
> a similar one?
> >
> > Thanks,
> > Marcel
> >
> > --
> > ImageJ mailing list: http://imagej.nih.gov/ij/list.html
>
> --
> ImageJ mailing list: http://imagej.nih.gov/ij/list.html
>

--
ImageJ mailing list: http://imagej.nih.gov/ij/list.html

[Jean-Louis Oneto]

In your specific case (sunset), I think you can make the following safe
assumptions:
1) most (if not all) of the phenomena comes from Rayleigh scattering
through the atmosphere
2) there is no reason that strong discrete spectral lines appears in the
resulting spectra (the diverse absorption line should be negligible here)
2') I take for granted that you avoid strange (and rare) weather
phenomenae like green light, boreal aurors...
3) the main part of the result should be close to a shift in color
temperature, the Sun light (outside atmosphere) being very close to a
black body radiation.
Then, why don't you try to just make a White Balance in order to
estimate the color temperature of your pictures? You should calibrate
your camera first (with some pictures of calibration targets taken under
controlled lighting), but after that several image processing programs
(I use Rawtherapee for that, I'm not sure about ImageJ) will give you
directly a measure of the color temperature, that you can translate in
wavelength of the maximum black body radiation power curve using the
Wien displacement law:

    \lambda_\max = \frac{b}{T}

where the constant,/b/, known as Wien's displacement constant, is equal
to2.8977721(26)×10^−3  K m.^[32]
<http://en.wikipedia.org/wiki/Black-body_radiation#cite_note-32>
[quoted from Wikipedia: http://en.wikipedia.org/wiki/Black-body_radiation ]

Best regards,
Jean-Louis


On 30/01/2015 01:17, Marcel Tschudin wrote:

> 1) Sorry, I made a big mistake by designating the balance point in the
> response spectrum of a detector "dominant wavelength". I was not aware that
> this expression is already used in colour science. I noticed also that
> others tend to misinterpret it as the statistical modal value. I find
> "balance wavelength" is less confusing and relates even better to the
> initial purpose, but there might even be better ones.
>
> 2) I hoped to obtain from the RGB values of the sun image, which
> approximate the perceived colour of it, an estimate for the "balance
> wavelength" of the camera's response spectrum. The sun's spectrum outside
> the atmosphere is known. With a fairly high probability the result will be
> between e.g. 500nm and 700nm. The transmission through the atmosphere can
> be estimated using aerosol measurements. But here we start now to estimate
> the "balance wavelength" by simulation, and I actually wondered whether I
> could obtain from the photos an estimate independent of the simulation.
>
> Marcel
>
> On Wed, Jan 28, 2015 at 6:34 AM, Kenneth Sloan <[hidden email]>
> wrote:
>
>> How many assumptions are you willing to make?
>>
>> The human eye is no more able to compute the “dominant wavelength” of an
>> arbitrary spectrum than a camera is.  But, the human eye did not evolve in
>> an environment of arbitrary spectra.
>>
>> If you assume “natural light”, or “black body radiation”, then there is
>> some hope.  If you want to do this in the face of arbitrary lighting
>> conditions - then, no - it simply can’t be done.  Either by the human
>> visual system or a traditional camera.
>>
>> How many samples are required (across the spectrum)?  Alas, in general the
>> answer is: an infinite number. Any claim that you can use fewer is simply
>> an assertion about the nature of the lighting.
>>
>> In fact, even the concept of “Dominant Wavelength” involves considerable
>> assumptions.  In fact…in those cases where “Dominant Wavelength” makes
>> sense… then human eyes and conventional RGB cameras can do the job just
>> fine.  The difficulty comes when you try to extend this concept into
>> domains where it doesn’t apply.  “Dominant Wavelength” is a concept that
>> really only makes sense in a color system that is 3D - one where “Hue” is a
>> possible dimension.  It’s really a psychological concept, not a physical
>> one [except to the extent that psychology evolved to match a certain flavor
>> of physical reality].
>>
>> Confused, yet?  Good!
>>
>> --
>> Kenneth Sloan
>> [hidden email]
>>
>>
>>> On Jan 27, 2015, at 13:35 , Marcel Tschudin <[hidden email]>
>> wrote:
>>> Hello everyone,
>>>
>>> I am new here. I am wondering whether I could use ImageJ (or an other
>> program) for measuring in photos the 'Dominant Wavelength' of the colors
>> within a selected pixel area. I provide here some further explanations
>> because I am not sure whether what I intend to do would actually even be
>> possible with photos.
>>> I would like to estimate the sun's 'Dominant Wavelength' in photos of
>> the setting sun. For a detector like the human eye the 'Dominant
>> Wavelength' would result from the sun's spectrum after passing the
>> atmosphere and after passing the eye's spectral detector efficiency. It
>> would be calculated from the detector's spectrum as Ldom, with the
>> radiation intensity, I, at a certain wavelength, L, in increments, dL, over
>> the visible spectrum as a ratio of two sums (integrals):
>>> Ldom = Sum(I*L*dL) / Sum(I*dL)
>>> (Because 'Dominant Wavelength' could be misinterpreted others suggest to
>> call this the 'Balanced Wavelength' instead.)
>>> Consumer cameras do not record the spectrum, they rather approximate the
>> detected spectral content, i.e. the color perceived by the human eye, with
>> the RGB information. Would it now be possible to estimate the original
>> 'Dominant Wavelength' from the available RGB information in the photo? If
>> yes, do you know if ImageJ (or an other program) provides such a feature or
>> a similar one?
>>> Thanks,
>>> Marcel
>>>
>>> --
>>> ImageJ mailing list: http://imagej.nih.gov/ij/list.html
>> --
>> ImageJ mailing list: http://imagej.nih.gov/ij/list.html
>>
> --
> ImageJ mailing list: http://imagej.nih.gov/ij/list.html
>

--
Jean-Louis Oneto
email: [hidden email]


--
ImageJ mailing list: http://imagej.nih.gov/ij/list.html

[Jean-Louis Oneto]

[second try, it seems that just the picture of the equation went through
the list (I would have rather suspected the opposite !)]
Hello,
In your specific case (sunset), I think you can make the following safe
assumptions:
1) most (if not all) of the phenomena comes from Rayleigh scattering
through the atmosphere
2) there is no reason that strong discrete spectral lines appears in the
resulting spectra (the diverse absorption line should be negligible here)
2') I take for granted that you avoid strange (and rare) weather
phenomenae like green light, boreal aurors...
3) the main part of the result should be close to a shift in color
temperature, the Sun light (outside atmosphere) being very close to a
black body radiation.
Then, why don't you try to just make a White Balance in order to
estimate the color temperature of your pictures? You should calibrate
your camera first (with some pictures of calibration targets taken under
controlled lighting), but after that several image processing programs
(I use Rawtherapee for that, I'm not sure about ImageJ) will give you
directly a measure of the color temperature, that you can translate in
wavelength of the maximum black body radiation power curve using the
Wien displacement law:

    \lambda_\max = \frac{b}{T}

where the constant,/b/, known as Wien's displacement constant, is equal
to2.8977721(26)×10^−3  K m.^[32]
[quoted from Wikipedia: http://en.wikipedia.org/wiki/Black-body_radiation ]

Best regards,
Jean-Louis

On 30/01/2015 01:17, Marcel Tschudin wrote:

> 1) Sorry, I made a big mistake by designating the balance point in the
> response spectrum of a detector "dominant wavelength". I was not aware that
> this expression is already used in colour science. I noticed also that
> others tend to misinterpret it as the statistical modal value. I find
> "balance wavelength" is less confusing and relates even better to the
> initial purpose, but there might even be better ones.
>
> 2) I hoped to obtain from the RGB values of the sun image, which
> approximate the perceived colour of it, an estimate for the "balance
> wavelength" of the camera's response spectrum. The sun's spectrum outside
> the atmosphere is known. With a fairly high probability the result will be
> between e.g. 500nm and 700nm. The transmission through the atmosphere can
> be estimated using aerosol measurements. But here we start now to estimate
> the "balance wavelength" by simulation, and I actually wondered whether I
> could obtain from the photos an estimate independent of the simulation.
>
> Marcel
>
> On Wed, Jan 28, 2015 at 6:34 AM, Kenneth Sloan <[hidden email]>
> wrote:
>
>> How many assumptions are you willing to make?
>>
>> The human eye is no more able to compute the “dominant wavelength” of an
>> arbitrary spectrum than a camera is.  But, the human eye did not evolve in
>> an environment of arbitrary spectra.
>>
>> If you assume “natural light”, or “black body radiation”, then there is
>> some hope.  If you want to do this in the face of arbitrary lighting
>> conditions - then, no - it simply can’t be done.  Either by the human
>> visual system or a traditional camera.
>>
>> How many samples are required (across the spectrum)?  Alas, in general the
>> answer is: an infinite number. Any claim that you can use fewer is simply
>> an assertion about the nature of the lighting.
>>
>> In fact, even the concept of “Dominant Wavelength” involves considerable
>> assumptions.  In fact…in those cases where “Dominant Wavelength” makes
>> sense… then human eyes and conventional RGB cameras can do the job just
>> fine.  The difficulty comes when you try to extend this concept into
>> domains where it doesn’t apply.  “Dominant Wavelength” is a concept that
>> really only makes sense in a color system that is 3D - one where “Hue” is a
>> possible dimension.  It’s really a psychological concept, not a physical
>> one [except to the extent that psychology evolved to match a certain flavor
>> of physical reality].
>>
>> Confused, yet?  Good!
>>
>> --
>> Kenneth Sloan
>> [hidden email]
>>
>>
>>> On Jan 27, 2015, at 13:35 , Marcel Tschudin <[hidden email]>
>> wrote:
>>> Hello everyone,
>>>
>>> I am new here. I am wondering whether I could use ImageJ (or an other
>> program) for measuring in photos the 'Dominant Wavelength' of the colors
>> within a selected pixel area. I provide here some further explanations
>> because I am not sure whether what I intend to do would actually even be
>> possible with photos.
>>> I would like to estimate the sun's 'Dominant Wavelength' in photos of
>> the setting sun. For a detector like the human eye the 'Dominant
>> Wavelength' would result from the sun's spectrum after passing the
>> atmosphere and after passing the eye's spectral detector efficiency. It
>> would be calculated from the detector's spectrum as Ldom, with the
>> radiation intensity, I, at a certain wavelength, L, in increments, dL, over
>> the visible spectrum as a ratio of two sums (integrals):
>>> Ldom = Sum(I*L*dL) / Sum(I*dL)
>>> (Because 'Dominant Wavelength' could be misinterpreted others suggest to
>> call this the 'Balanced Wavelength' instead.)
>>> Consumer cameras do not record the spectrum, they rather approximate the
>> detected spectral content, i.e. the color perceived by the human eye, with
>> the RGB information. Would it now be possible to estimate the original
>> 'Dominant Wavelength' from the available RGB information in the photo? If
>> yes, do you know if ImageJ (or an other program) provides such a feature or
>> a similar one?
>>> Thanks,
>>> Marcel
>>>
>>> --
>>> ImageJ mailing list: http://imagej.nih.gov/ij/list.html
>> --
>> ImageJ mailing list: http://imagej.nih.gov/ij/list.html
>>
> --
> ImageJ mailing list: http://imagej.nih.gov/ij/list.html
>

--
Jean-Louis Oneto
email: [hidden email]


--
ImageJ mailing list: http://imagej.nih.gov/ij/list.html

[Marcel Tschudin]

@Jean-Louis
Thank you for your suggestion to estimate the wavelength with the colour
temperature. That should indeed be possible provided I can calibrate the
pocket zoom camera which was used with the intention to measure only angles.
Cordialement,
Marcel

On Fri, Jan 30, 2015 at 5:53 AM, Jean-Louis Oneto <[hidden email]> wrote:

> In your specific case (sunset), I think you can make the following safe
> assumptions:
> 1) most (if not all) of the phenomena comes from Rayleigh scattering
> through the atmosphere
> 2) there is no reason that strong discrete spectral lines appears in the
> resulting spectra (the diverse absorption line should be negligible here)
> 2') I take for granted that you avoid strange (and rare) weather
> phenomenae like green light, boreal aurors...
> 3) the main part of the result should be close to a shift in color
> temperature, the Sun light (outside atmosphere) being very close to a black
> body radiation.
> Then, why don't you try to just make a White Balance in order to estimate
> the color temperature of your pictures? You should calibrate your camera
> first (with some pictures of calibration targets taken under controlled
> lighting), but after that several image processing programs (I use
> Rawtherapee for that, I'm not sure about ImageJ) will give you directly a
> measure of the color temperature, that you can translate in wavelength of
> the maximum black body radiation power curve using the Wien displacement
> law:
>
>    \lambda_\max = \frac{b}{T}
>
> where the constant,/b/, known as Wien's displacement constant, is equal
> to2.8977721(26)×10^−3  K m.^[32] <http://en.wikipedia.org/wiki/
> Black-body_radiation#cite_note-32>
> [quoted from Wikipedia: http://en.wikipedia.org/wiki/Black-body_radiation
> ]
>
> Best regards,
> Jean-Louis
>
>
>
> On 30/01/2015 01:17, Marcel Tschudin wrote:
>
>> 1) Sorry, I made a big mistake by designating the balance point in the
>> response spectrum of a detector "dominant wavelength". I was not aware
>> that
>> this expression is already used in colour science. I noticed also that
>> others tend to misinterpret it as the statistical modal value. I find
>> "balance wavelength" is less confusing and relates even better to the
>> initial purpose, but there might even be better ones.
>>
>> 2) I hoped to obtain from the RGB values of the sun image, which
>> approximate the perceived colour of it, an estimate for the "balance
>> wavelength" of the camera's response spectrum. The sun's spectrum outside
>> the atmosphere is known. With a fairly high probability the result will be
>> between e.g. 500nm and 700nm. The transmission through the atmosphere can
>> be estimated using aerosol measurements. But here we start now to estimate
>> the "balance wavelength" by simulation, and I actually wondered whether I
>> could obtain from the photos an estimate independent of the simulation.
>>
>> Marcel
>>
>> On Wed, Jan 28, 2015 at 6:34 AM, Kenneth Sloan <[hidden email]>
>> wrote:
>>
>>  How many assumptions are you willing to make?
>>>
>>> The human eye is no more able to compute the “dominant wavelength” of an
>>> arbitrary spectrum than a camera is.  But, the human eye did not evolve
>>> in
>>> an environment of arbitrary spectra.
>>>
>>> If you assume “natural light”, or “black body radiation”, then there is
>>> some hope.  If you want to do this in the face of arbitrary lighting
>>> conditions - then, no - it simply can’t be done.  Either by the human
>>> visual system or a traditional camera.
>>>
>>> How many samples are required (across the spectrum)?  Alas, in general
>>> the
>>> answer is: an infinite number. Any claim that you can use fewer is simply
>>> an assertion about the nature of the lighting.
>>>
>>> In fact, even the concept of “Dominant Wavelength” involves considerable
>>> assumptions.  In fact…in those cases where “Dominant Wavelength” makes
>>> sense… then human eyes and conventional RGB cameras can do the job just
>>> fine.  The difficulty comes when you try to extend this concept into
>>> domains where it doesn’t apply.  “Dominant Wavelength” is a concept that
>>> really only makes sense in a color system that is 3D - one where “Hue”
>>> is a
>>> possible dimension.  It’s really a psychological concept, not a physical
>>> one [except to the extent that psychology evolved to match a certain
>>> flavor
>>> of physical reality].
>>>
>>> Confused, yet?  Good!
>>>
>>> --
>>> Kenneth Sloan
>>> [hidden email]
>>>
>>>
>>>  On Jan 27, 2015, at 13:35 , Marcel Tschudin <
>>>> [hidden email]>
>>>>
>>> wrote:
>>>
>>>> Hello everyone,
>>>>
>>>> I am new here. I am wondering whether I could use ImageJ (or an other
>>>>
>>> program) for measuring in photos the 'Dominant Wavelength' of the colors
>>> within a selected pixel area. I provide here some further explanations
>>> because I am not sure whether what I intend to do would actually even be
>>> possible with photos.
>>>
>>>> I would like to estimate the sun's 'Dominant Wavelength' in photos of
>>>>
>>> the setting sun. For a detector like the human eye the 'Dominant
>>> Wavelength' would result from the sun's spectrum after passing the
>>> atmosphere and after passing the eye's spectral detector efficiency. It
>>> would be calculated from the detector's spectrum as Ldom, with the
>>> radiation intensity, I, at a certain wavelength, L, in increments, dL,
>>> over
>>> the visible spectrum as a ratio of two sums (integrals):
>>>
>>>> Ldom = Sum(I*L*dL) / Sum(I*dL)
>>>> (Because 'Dominant Wavelength' could be misinterpreted others suggest to
>>>>
>>> call this the 'Balanced Wavelength' instead.)
>>>
>>>> Consumer cameras do not record the spectrum, they rather approximate the
>>>>
>>> detected spectral content, i.e. the color perceived by the human eye,
>>> with
>>> the RGB information. Would it now be possible to estimate the original
>>> 'Dominant Wavelength' from the available RGB information in the photo? If
>>> yes, do you know if ImageJ (or an other program) provides such a feature
>>> or
>>> a similar one?
>>>
>>>> Thanks,
>>>> Marcel
>>>>
>>>> --
>>>> ImageJ mailing list: http://imagej.nih.gov/ij/list.html
>>>>
>>> --
>>> ImageJ mailing list: http://imagej.nih.gov/ij/list.html
>>>
>>>  --
>> ImageJ mailing list: http://imagej.nih.gov/ij/list.html
>>
>>
> --
> Jean-Louis Oneto
> email: [hidden email]
>
>
>
> --
> ImageJ mailing list: http://imagej.nih.gov/ij/list.html
>

--
ImageJ mailing list: http://imagej.nih.gov/ij/list.html