Extract source spectrum from JPEG?

Howdy all!

To what extent is it possible to extract the `spectrum' of a light source
from, e.g., a digital camera JPEG, with existing freely-distributed software,
such as the Gimp? Can someone point me to such a tool?
(I've googled this group's archives but found nothing specific.)

EG: I'd like to demonstrate that the newish green traffic lights
using GaN-based semiconductors are more spectrally pure than the old
incandescent light/filter combinations they replace. I intend to
(i) deliberately defocus the camera, (ii) photograph each light source
(iii) identify the spectral intensity of each light source and compare.

Of course, this is vaguely what we are doing each time we *take*
a digital photograph: from an unknown subject taken through 3 different
color filters, we let the camera firmware reconstitute the full
color content of the original scene. The only difference is that I
want explicit information about the relative abundances of each spectral
color in the form of a continuous spectrum, ranging over the visible spectrum.
The R, G, and B intensity curves are only slightly useful, alas, and of
course the DETAILS of the spectral intensity probably require knowledge
of the details of the CCDs used to acquire the image...

Many thanks!

--
David WOOD
Laboratoire d'Optique des Solides
; 01.44.27.45.07
On sabbatic Jun-Dec 2004 from Dept. of Physics, Colorado School of Mines:

David M. Wood wrote:

Howdy all!

To what extent is it possible to extract the `spectrum' of a light source
from, e.g., a digital camera JPEG, with existing freely-distributed software,
such as the Gimp? Can someone point me to such a tool?
(I've googled this group's archives but found nothing specific.)

Not a price. The only way to do it is to photograph the spectrum of the
light diffracted off a grating or refracted through a prism. Shovelware
CDs (use an aluminium one) provide cheap reflection diffraction
gratings. Or you could buy a Cokin 2 point spectrum star filter.

EG: I'd like to demonstrate that the newish green traffic lights
using GaN-based semiconductors are more spectrally pure than the old
incandescent light/filter combinations they replace. I intend to
(i) deliberately defocus the camera, (ii) photograph each light source
(iii) identify the spectral intensity of each light source and compare.

Of course, this is vaguely what we are doing each time we *take*
a digital photograph: from an unknown subject taken through 3 different
color filters, we let the camera firmware reconstitute the full
color content of the original scene. The only difference is that I
want explicit information about the relative abundances of each spectral
color in the form of a continuous spectrum, ranging over the visible spectrum.

The only way to do that is separate the light by wavelength before it
hits the sensor. Examples on streetlamps at:

http://www.nezumi.demon.co.uk/nonad/spectra.htm

Conventional coloured glass filters are very impure colourer light. LED
technology is typically outputs most power in about 50nm bandwidth with
a tail.

Regards,
Martin Brown

David M. Wood wrote:

Howdy all!

To what extent is it possible to extract the `spectrum' of a light source
from, e.g., a digital camera JPEG, with existing freely-distributed software,
such as the Gimp? Can someone point me to such a tool?
(I've googled this group's archives but found nothing specific.)

Not a price. The only way to do it is to photograph the spectrum of the
light diffracted off a grating or refracted through a prism. Shovelware
CDs (use an aluminium one) provide cheap reflection diffraction
gratings. Or you could buy a Cokin 2 point spectrum star filter.

EG: I'd like to demonstrate that the newish green traffic lights
using GaN-based semiconductors are more spectrally pure than the old
incandescent light/filter combinations they replace. I intend to
(i) deliberately defocus the camera, (ii) photograph each light source
(iii) identify the spectral intensity of each light source and compare.

Of course, this is vaguely what we are doing each time we *take*
a digital photograph: from an unknown subject taken through 3 different
color filters, we let the camera firmware reconstitute the full
color content of the original scene. The only difference is that I
want explicit information about the relative abundances of each spectral
color in the form of a continuous spectrum, ranging over the visible spectrum.

The only way to do that is separate the light by wavelength before it
hits the sensor. Examples on streetlamps at:

http://www.nezumi.demon.co.uk/nonad/spectra.htm

Conventional coloured glass filters are very impure colourer light. LED
technology is typically outputs most power in about 50nm bandwidth with
a tail.

Regards,
Martin Brown

(David M. Wood) writes:

Howdy all!

To what extent is it possible to extract the `spectrum' of a light source
from, e.g., a digital camera JPEG, with existing freely-distributed software,
such as the Gimp? Can someone point me to such a tool?
(I've googled this group's archives but found nothing specific.)

In any more than a trivial sense, impossible.

EG: I'd like to demonstrate that the newish green traffic lights
using GaN-based semiconductors are more spectrally pure than the old
incandescent light/filter combinations they replace. I intend to
(i) deliberately defocus the camera, (ii) photograph each light source
(iii) identify the spectral intensity of each light source and compare.

You're missing the point here. This isn't going to work.

Of course, this is vaguely what we are doing each time we *take*
a digital photograph: from an unknown subject taken through 3 different
color filters, we let the camera firmware reconstitute the full
color content of the original scene. The only difference is that I
want explicit information about the relative abundances of each spectral
color in the form of a continuous spectrum, ranging over the visible spectrum.
The R, G, and B intensity curves are only slightly useful, alas, and of
course the DETAILS of the spectral intensity probably require knowledge
of the details of the CCDs used to acquire the image...

There are not '3 different spectral colours'. Visible light lives in
about a 300+nm wide band. A 'spectrally pure' light source is going to
be less than 1nm wide.

Now, break that 350-odd nm into three slightly overlapping chunks,
convolve an odd hump over each one and tell me precisely how you think
you're going to make any useful observations about spectral purity
with the final area-under-function integral numbers.

If you want to do this, go and buy a spectrometer. They aren't
particularly expensive and they'll do precisely what you want.

B

(David M. Wood) writes:

Howdy all!

To what extent is it possible to extract the `spectrum' of a light source
from, e.g., a digital camera JPEG, with existing freely-distributed software,
such as the Gimp? Can someone point me to such a tool?
(I've googled this group's archives but found nothing specific.)

In any more than a trivial sense, impossible.

EG: I'd like to demonstrate that the newish green traffic lights
using GaN-based semiconductors are more spectrally pure than the old
incandescent light/filter combinations they replace. I intend to
(i) deliberately defocus the camera, (ii) photograph each light source
(iii) identify the spectral intensity of each light source and compare.

You're missing the point here. This isn't going to work.

Of course, this is vaguely what we are doing each time we *take*
a digital photograph: from an unknown subject taken through 3 different
color filters, we let the camera firmware reconstitute the full
color content of the original scene. The only difference is that I
want explicit information about the relative abundances of each spectral
color in the form of a continuous spectrum, ranging over the visible spectrum.
The R, G, and B intensity curves are only slightly useful, alas, and of
course the DETAILS of the spectral intensity probably require knowledge
of the details of the CCDs used to acquire the image...

There are not '3 different spectral colours'. Visible light lives in
about a 300+nm wide band. A 'spectrally pure' light source is going to
be less than 1nm wide.

Now, break that 350-odd nm into three slightly overlapping chunks,
convolve an odd hump over each one and tell me precisely how you think
you're going to make any useful observations about spectral purity
with the final area-under-function integral numbers.

If you want to do this, go and buy a spectrometer. They aren't
particularly expensive and they'll do precisely what you want.

B

David M. Wood wrote:
....
EG: I'd like to demonstrate that the newish green traffic lights
using GaN-based semiconductors are more spectrally pure than the old
incandescent light/filter combinations they replace. I intend to
(i) deliberately defocus the camera, (ii) photograph each light source
(iii) identify the spectral intensity of each light source and compare.

This has *got* to be some governmental thing.
Only governments could find a very complicated way to find out if a
green light is green!
How much simpler would it be to just put a few people in front of the
light, and ask, "Does this look green to you?"
The actual spectrum purity wouldn't matter. It would eithe rlook green
or it wouldn't. yes, it could be altered by adding blue or yellow
light, but that's easily done, too.
Oh well, it's just my taxes. :-)

--
Bill Funk
Change "g" to "a"

Rent or borrow a spectrophotometer, and measure them directly.

[Edmund Optics has one at a *relatively* low cost ...
http://www.edmundoptics.com/onlineca...productID=2306
&search=1


As "an aside" ... it's the "scientifically correct" way of doing this task,
and it's probably the only way that "would stand up in court."








"David M. Wood" wrote in message
...
Howdy all!

To what extent is it possible to extract the `spectrum' of a light source
from, e.g., a digital camera JPEG, with existing freely-distributed
software,
such as the Gimp? Can someone point me to such a tool?
(I've googled this group's archives but found nothing specific.)

EG: I'd like to demonstrate that the newish green traffic lights
using GaN-based semiconductors are more spectrally pure than the old
incandescent light/filter combinations they replace. I intend to
(i) deliberately defocus the camera, (ii) photograph each light source
(iii) identify the spectral intensity of each light source and compare.

Of course, this is vaguely what we are doing each time we *take*
a digital photograph: from an unknown subject taken through 3 different
color filters, we let the camera firmware reconstitute the full
color content of the original scene. The only difference is that I
want explicit information about the relative abundances of each spectral
color in the form of a continuous spectrum, ranging over the visible
spectrum.
The R, G, and B intensity curves are only slightly useful, alas, and of
course the DETAILS of the spectral intensity probably require knowledge
of the details of the CCDs used to acquire the image...

Many thanks!

--
David WOOD
Laboratoire d'Optique des Solides
; 01.44.27.45.07
On sabbatic Jun-Dec 2004 from Dept. of Physics, Colorado School of Mines:

Rent or borrow a spectrophotometer, and measure them directly.

[Edmund Optics has one at a *relatively* low cost ...
http://www.edmundoptics.com/onlineca...productID=2306
&search=1


As "an aside" ... it's the "scientifically correct" way of doing this task,
and it's probably the only way that "would stand up in court."








"David M. Wood" wrote in message
...
Howdy all!

To what extent is it possible to extract the `spectrum' of a light source
from, e.g., a digital camera JPEG, with existing freely-distributed
software,
such as the Gimp? Can someone point me to such a tool?
(I've googled this group's archives but found nothing specific.)

EG: I'd like to demonstrate that the newish green traffic lights
using GaN-based semiconductors are more spectrally pure than the old
incandescent light/filter combinations they replace. I intend to
(i) deliberately defocus the camera, (ii) photograph each light source
(iii) identify the spectral intensity of each light source and compare.

Of course, this is vaguely what we are doing each time we *take*
a digital photograph: from an unknown subject taken through 3 different
color filters, we let the camera firmware reconstitute the full
color content of the original scene. The only difference is that I
want explicit information about the relative abundances of each spectral
color in the form of a continuous spectrum, ranging over the visible
spectrum.
The R, G, and B intensity curves are only slightly useful, alas, and of
course the DETAILS of the spectral intensity probably require knowledge
of the details of the CCDs used to acquire the image...

Many thanks!

--
David WOOD
Laboratoire d'Optique des Solides
; 01.44.27.45.07
On sabbatic Jun-Dec 2004 from Dept. of Physics, Colorado School of Mines:

Big Bill writes:

David M. Wood wrote:
...
EG: I'd like to demonstrate that the newish green traffic lights
using GaN-based semiconductors are more spectrally pure than the old
incandescent light/filter combinations they replace. I intend to
(i) deliberately defocus the camera, (ii) photograph each light source
(iii) identify the spectral intensity of each light source and compare.

This has *got* to be some governmental thing.
Only governments could find a very complicated way to find out if a
green light is green!

Governments or people with a clue, which you appear not to have.

How much simpler would it be to just put a few people in front of the
light, and ask, "Does this look green to you?"

Since people perceive colour in a variety of ways, that's a stupid
bloody idea. I know! let's fly in the face of many decades of colour
theory!

The actual spectrum purity wouldn't matter. It would eithe rlook green
or it wouldn't. yes, it could be altered by adding blue or yellow
light, but that's easily done, too.

Actually, the spectral purity *does* matter, particularly when you've
got people with limit colour vision around who should not have
suddenly unexpected results.

B

Big Bill writes:

David M. Wood wrote:
...
EG: I'd like to demonstrate that the newish green traffic lights
using GaN-based semiconductors are more spectrally pure than the old
incandescent light/filter combinations they replace. I intend to
(i) deliberately defocus the camera, (ii) photograph each light source
(iii) identify the spectral intensity of each light source and compare.

This has *got* to be some governmental thing.
Only governments could find a very complicated way to find out if a
green light is green!

Governments or people with a clue, which you appear not to have.

How much simpler would it be to just put a few people in front of the
light, and ask, "Does this look green to you?"

Since people perceive colour in a variety of ways, that's a stupid
bloody idea. I know! let's fly in the face of many decades of colour
theory!

The actual spectrum purity wouldn't matter. It would eithe rlook green
or it wouldn't. yes, it could be altered by adding blue or yellow
light, but that's easily done, too.

Actually, the spectral purity *does* matter, particularly when you've
got people with limit colour vision around who should not have
suddenly unexpected results.

B