How are digital cameras able to render violet?

Discussion in 'Digital Photography' started by Paul Ciszek, Feb 11, 2013.

  1. Paul Ciszek

    Paul Ciszek Guest

    The violet at the inner edge of the rainbow (outer edge of the second
    bow in a double) is not "purple", i.e., it is not a mixture of red and
    blue light, it is a pure spectral color in its own right--a wavelength
    shorter than blue. The human eye sees violet as looking purplish because
    our color vision is a bit of kludge--we detect "red" by subtracting a
    narrow "green" response from a wider, yellower "green" response and
    calling the difference "red". As an artifact of this subtraction, short
    wavelength violet light (which weakly triggers both types of green
    response) registers as a little bit "red" as well, while also triggering
    the blue response as you would expect. Ergo, violet light looks reddish-
    blue to humans. But why does it look redish-blue to a digital camera?
    Don't the sensors use actual red filters to detect red, unlike the human
    eye?

    I note that even old, cheap digital cameras manage it:
    http://www.flickr.com/photos/35853148@N05/4303188267

    --
    Please reply to: | No nation is drunken where wine is cheap.
    pciszek at panix dot com | --Thomas Jefferson
    Paul Ciszek, Feb 11, 2013
    #1
    1. Advertising

  2. Paul Ciszek

    RichA Guest

    On Feb 11, 4:15 pm, (Paul Ciszek) wrote:
    > The violet at the inner edge of the rainbow (outer edge of the second
    > bow in a double) is not "purple", i.e., it is not a mixture of red and
    > blue light, it is a pure spectral color in its own right--a wavelength
    > shorter than blue. The human eye sees violet as looking purplish because
    > our color vision is a bit of kludge--we detect "red" by subtracting a
    > narrow "green" response from a wider, yellower "green" response and
    > calling the difference "red". As an artifact of this subtraction, short
    > wavelength violet light (which weakly triggers both types of green
    > response) registers as a little bit "red" as well, while also triggering
    > the blue response as you would expect. Ergo, violet light looks reddish-
    > blue to humans. But why does it look redish-blue to a digital camera?
    > Don't the sensors use actual red filters to detect red, unlike the human
    > eye?
    >
    > I note that even old, cheap digital cameras manage it:http://www.flickr.com/photos/35853148@N05/4303188267
    >



    Human vision extends into the violet (about 1% efficiency) and can see
    it. CMOS cameras are about 50% efficient in the violet and probably
    renders it by combining red and blue filter elements.
    There is no such thing as a reddish-blue colour, at least not
    practically, since you cannot resolve the photons of red and blue
    reflections so violet is what you see.
    RichA, Feb 12, 2013
    #2
    1. Advertising

  3. Paul Ciszek

    Paul Ciszek Guest

    In article <>,
    RichA <> wrote:
    >
    >Human vision extends into the violet (about 1% efficiency) and can see
    >it. CMOS cameras are about 50% efficient in the violet and probably
    >renders it by combining red and blue filter elements.


    Being sensetive to a wavelength of light, and being able to distinguish it
    as a color are two different things. B&W film responds to many colors of
    light, but doesn't differentiate them. Color film used to respond to UV,
    but it would just show up as blue. IR response was weird.

    In order for a digital camera to render violet light as purplish, as in
    the rainbow picture I linked to, both the red pixels and the blue pixels
    would have to be responding to violet light. I understand why this
    happens in the human eye--as I explained, it's because we don't have a
    separate "red" detector as such--but in order for the bayer filter of a
    color digital camera to render violet as purple rather than blue, the
    "red" filters would have to pass the long wavelengths, block the green
    and blue wavelengths, and pass again at the very shortest wavelengths.
    This is not easy to do, especially if your filter has to be deposited
    onto a semiconductor surface rather than mixed up in a vat by Schott
    Glass.

    >There is no such thing as a reddish-blue colour, at least not
    >practically, since you cannot resolve the photons of red and blue
    >reflections so violet is what you see.


    There most certainly is such a thing as "reddish blue". Any purple you
    see printed, rendered on a monitor, or projected on a movie screen
    contains both red and blue light. Violet light exists in nature and
    can be produced by some light sources, but any form of media fakes it
    using purple--aka "reddish blue"--instead. Ironically, so do violets--
    they are purple rather than truely violet.

    --
    Please reply to: | "We establish no religion in this country, we
    pciszek at panix dot com | command no worship, we mandate no belief, nor
    Autoreply is disabled | will we ever. Church and state are, and must
    | remain, separate." --Ronald Reagan, 10/26/1984
    Paul Ciszek, Feb 12, 2013
    #3
  4. On 2/12/2013 6:12 AM, Paul Ciszek wrote:
    > In article <>,
    > RichA <> wrote:
    >> Human vision extends into the violet (about 1% efficiency) and can see
    >> it. CMOS cameras are about 50% efficient in the violet and probably
    >> renders it by combining red and blue filter elements.

    > Being sensetive to a wavelength of light, and being able to distinguish it
    > as a color are two different things. B&W film responds to many colors of
    > light, but doesn't differentiate them. Color film used to respond to UV,
    > but it would just show up as blue. IR response was weird.
    >
    > In order for a digital camera to render violet light as purplish, as in
    > the rainbow picture I linked to, both the red pixels and the blue pixels
    > would have to be responding to violet light. I understand why this
    > happens in the human eye--as I explained, it's because we don't have a
    > separate "red" detector as such--but in order for the bayer filter of a
    > color digital camera to render violet as purple rather than blue, the
    > "red" filters would have to pass the long wavelengths, block the green
    > and blue wavelengths, and pass again at the very shortest wavelengths.
    > This is not easy to do, especially if your filter has to be deposited
    > onto a semiconductor surface rather than mixed up in a vat by Schott
    > Glass.
    >
    >> There is no such thing as a reddish-blue colour, at least not
    >> practically, since you cannot resolve the photons of red and blue
    >> reflections so violet is what you see.

    > There most certainly is such a thing as "reddish blue". Any purple you
    > see printed, rendered on a monitor, or projected on a movie screen
    > contains both red and blue light. Violet light exists in nature and
    > can be produced by some light sources, but any form of media fakes it
    > using purple--aka "reddish blue"--instead. Ironically, so do violets--
    > they are purple rather than truely violet.
    >

    I remember deep blue gentians in the Swiss mountains being reproduced as
    pink using Kodachrome and I was interested to get the same effect
    photographing deep blue petunias with a digital camera. Both cases are
    due to high reflectivity of UV light by the flowers.

    --
    Jim Silverton (Potomac, MD)

    Extraneous "not" in Reply To.
    James Silverton, Feb 12, 2013
    #4
  5. Paul Ciszek

    Me Guest

    On 12/02/2013 10:15 a.m., Paul Ciszek wrote:
    > The violet at the inner edge of the rainbow (outer edge of the second
    > bow in a double) is not "purple", i.e., it is not a mixture of red and
    > blue light, it is a pure spectral color in its own right--a wavelength
    > shorter than blue. The human eye sees violet as looking purplish because
    > our color vision is a bit of kludge--we detect "red" by subtracting a
    > narrow "green" response from a wider, yellower "green" response and
    > calling the difference "red". As an artifact of this subtraction, short
    > wavelength violet light (which weakly triggers both types of green
    > response) registers as a little bit "red" as well, while also triggering
    > the blue response as you would expect. Ergo, violet light looks reddish-
    > blue to humans. But why does it look redish-blue to a digital camera?
    > Don't the sensors use actual red filters to detect red, unlike the human
    > eye?
    >
    > I note that even old, cheap digital cameras manage it:
    > http://www.flickr.com/photos/35853148@N05/4303188267
    >

    The red filter in the Bayer array probably passes some violet light.
    Here are some transmission charts (for photographic filters).
    This red filters shows a (second) "transmission peak" at the violet end
    of the visible spectrum. Perhaps - more correctly - the chart should be
    inverted so you look at it as an absorption peak rather than
    transmission peak.
    >

    http://www.schneiderkreuznach.com/pdf/filter/bw_filter_transmission_curves.pdf
    (see page 2)

    The absorption spectrum of Bayer arrays won't be linear, neither will
    the spectral response of the photodetectors. This is adjusted in-camera
    by colour tone curve or in PP of raw files.
    In theory, Foveon sensors have higher quantum efficiency (than Bayer
    CFA), as the colour filter layers are the photodetector itself (not
    "wasting" photons in a separate colour filter array) - but the filtering
    is so imprecise that adjustment more than exceeds any theoretical gain,
    and practical QE is less than that achieved by Bayer CFA sensors.
    Me, Feb 12, 2013
    #5
  6. Paul Ciszek

    RichA Guest

    On Feb 12, 4:52 pm, jgh <> wrote:
    > On Tue, 12 Feb 2013 08:58:19 -0500, James Silverton wrote:
    > > I remember deep blue gentians in the Swiss mountains being reproduced as
    > > pink using Kodachrome and I was interested to get the same effect
    > > photographing deep blue petunias with a digital camera. Both cases are
    > > due to high reflectivity of UV light by the flowers.

    >
    > I've seen in-camera colour processing be confused by a violet; it dealt
    > with it successfully in "normal" mode but captured it as pink in "vivid
    > colour".
    >
    > This was a dyed wool colour, outside on a bright snowy winter's day.
    > --
    > Stick


    Violet isn't the major problem colour with sensors, green is.
    RichA, Feb 13, 2013
    #6
  7. Paul Ciszek

    Mort Guest

    James Silverton wrote:
    >>

    > I remember deep blue gentians in the Swiss mountains being reproduced as
    > pink using Kodachrome and I was interested to get the same effect
    > photographing deep blue petunias with a digital camera. Both cases are
    > due to high reflectivity of UV light by the flowers.



    Hi,

    I had a similar situation with morning glory flowers on Cape Cod, that
    were blue to the eye and came out pink on Kodachrome. When I shot them
    using Fujichrome, they came out the proper shade of blue, which greatly
    pleased the amateur horticulturist that I was visiting.

    I gave up long ago on aiming for "exact true color", and am satisfied in
    most cases with an approximation of the color. In some situations
    involving Medical or Scientific subjects, accurate color is required. Of
    course, now with digital photography, it is almost a moot point.

    Mort Linder
    Mort, Feb 13, 2013
    #7
  8. Paul Ciszek

    Paul Ciszek Guest

    In article <511aecd2$0$24758$>, Mort <> wrote:
    >
    >I gave up long ago on aiming for "exact true color", and am satisfied in
    >most cases with an approximation of the color. In some situations
    >involving Medical or Scientific subjects, accurate color is required. Of
    >course, now with digital photography, it is almost a moot point.


    Being digital doesn't make the problem go away. It just make it somewhat
    easier to fudge the results after the fact.


    --
    Please reply to: | "We establish no religion in this country, we
    pciszek at panix dot com | command no worship, we mandate no belief, nor
    Autoreply is disabled | will we ever. Church and state are, and must
    | remain, separate." --Ronald Reagan, 10/26/1984
    Paul Ciszek, Feb 13, 2013
    #8
  9. Paul Ciszek <> wrote:
    > The violet at the inner edge of the rainbow (outer edge of the second
    > bow in a double) is not "purple", i.e., it is not a mixture of red and
    > blue light, it is a pure spectral color in its own right--a wavelength
    > shorter than blue. The human eye sees violet as looking purplish because
    > our color vision is a bit of kludge--we detect "red" by subtracting a
    > narrow "green" response from a wider, yellower "green" response and
    > calling the difference "red". As an artifact of this subtraction, short
    > wavelength violet light (which weakly triggers both types of green
    > response) registers as a little bit "red" as well, while also triggering
    > the blue response as you would expect. Ergo, violet light looks reddish-
    > blue to humans. But why does it look redish-blue to a digital camera?
    > Don't the sensors use actual red filters to detect red, unlike the human
    > eye?


    > I note that even old, cheap digital cameras manage it:
    > http://www.flickr.com/photos/35853148@N05/4303188267


    They try hard to reproduce the colours as seen by the human
    eye/brain. Note they also don't distinguish between pure spectral
    green, and the faux green produced by a mixture of spectral blue and
    spectral yellow. It would be annoying if they did, since our
    eyes/brains can't tell the difference, but a goldfish probably could.

    In part it's an artefact of a three primary colour space, augmented by
    carefully choosing those filters which mostly give the best
    approximations to the colours we see. Probably the most obious problem
    area is the edge of spectrum pinks, blues, and violets of some
    flowers, especially if they're heavily oversaturated by fluorescence
    or have a high UV component because they're aiming at the eyes of
    insect with good UV colour sensing.
    Chris Malcolm, Feb 13, 2013
    #9
  10. Paul Ciszek

    Paul Ciszek Guest

    In article <kfedje$vr7$>, jgh <> wrote:
    >On Tue, 12 Feb 2013 08:58:19 -0500, James Silverton wrote:
    >> I remember deep blue gentians in the Swiss mountains being reproduced as
    >> pink using Kodachrome and I was interested to get the same effect
    >> photographing deep blue petunias with a digital camera. Both cases are
    >> due to high reflectivity of UV light by the flowers.

    >
    >I've seen in-camera colour processing be confused by a violet; it dealt
    >with it successfully in "normal" mode but captured it as pink in "vivid
    >colour".
    >
    >This was a dyed wool colour, outside on a bright snowy winter's day.


    I suspect that any fabric dye that appears "violet" is in fact "purple",
    i.e., reflects both red and blue light, just like the violet flower.

    --
    Please reply to: | "We establish no religion in this country, we
    pciszek at panix dot com | command no worship, we mandate no belief, nor
    Autoreply is disabled | will we ever. Church and state are, and must
    | remain, separate." --Ronald Reagan, 10/26/1984
    Paul Ciszek, Feb 13, 2013
    #10
  11. Paul Ciszek

    Paul Ciszek Guest

    In article <>,
    RichA <> wrote:
    >
    >Violet isn't the major problem colour with sensors, green is.


    I don't seem to have a problem with it:
    http://www.flickr.com/photos/35853148@N05/5929247242/


    --
    Please reply to: | "We establish no religion in this country, we
    pciszek at panix dot com | command no worship, we mandate no belief, nor
    Autoreply is disabled | will we ever. Church and state are, and must
    | remain, separate." --Ronald Reagan, 10/26/1984
    Paul Ciszek, Feb 13, 2013
    #11
  12. Paul Ciszek

    Paul Ciszek Guest

    In article <kfe8b3$sh6$>,
    Me <> wrote:
    >>

    >The red filter in the Bayer array probably passes some violet light.
    >Here are some transmission charts (for photographic filters).
    >This red filters shows a (second) "transmission peak" at the violet end
    >of the visible spectrum. Perhaps - more correctly - the chart should be
    >inverted so you look at it as an absorption peak rather than
    >transmission peak.
    > >

    >http://www.schneiderkreuznach.com/pdf/filter/bw_filter_transmission_curves.pdf
    >(see page 2)


    "Die von Ihnen gesuchte Seite existiert leider nicht oder der Link ist
    veraltet. Bitte verwenden Sie die Suche oder kehren Sie zur Startseite
    zurück, um die gewünschten Informationen zu finden."

    --
    Please reply to: | "We establish no religion in this country, we
    pciszek at panix dot com | command no worship, we mandate no belief, nor
    Autoreply is disabled | will we ever. Church and state are, and must
    | remain, separate." --Ronald Reagan, 10/26/1984
    Paul Ciszek, Feb 13, 2013
    #12
  13. Paul Ciszek

    Paul Ciszek Guest

    In article <>,
    Alfred Molon <> wrote:
    >In article <>, Chris Malcolm says...
    >> In part it's an artefact of a three primary colour space,

    >
    >Would the situation improve with four primary colours? Or more?


    I seem to remember reading that someon tried a red-yellow-green-blue
    mosaic filter, and it didn't help any. I'm not sure what you would
    actually do with the extra information. I suppose you could eventually
    work out improved versions of the different white balance modes, or the
    phony color filters that some cameras and post-processing programs are
    able to apply to RAW data.

    --
    Please reply to: | "We establish no religion in this country, we
    pciszek at panix dot com | command no worship, we mandate no belief, nor
    Autoreply is disabled | will we ever. Church and state are, and must
    | remain, separate." --Ronald Reagan, 10/26/1984
    Paul Ciszek, Feb 13, 2013
    #13
  14. Alfred Molon wrote:
    > In article <>, Chris Malcolm says...
    >> In part it's an artefact of a three primary colour space,

    >
    > Would the situation improve with four primary colours? Or more?


    Some women are tetrachromats, they have a rare mutation which gives them two
    copies of the 'red' receptor with different sensitivity peaks. IIRC it
    gives them slightly better discrimination in that end of the spectrum. The
    interesting thing is that the brain integrates this into a coherent picture
    of the world with no apparent effort.

    David
    David Hare-Scott, Feb 13, 2013
    #14
  15. On 2/13/2013 1:27 PM, Alfred Molon wrote:
    > In article <afde856a-b41a-4960-9ebb-
    > >, RichA says...
    >> Violet isn't the major problem colour with sensors, green is.

    > Why?

    I heard Edwin Land lecture on his two color theory and was totally
    convinced by his demonstrations. About the most spectacular was when he
    used the two sodium D-lines (both yellow) to reconstruct color. I don't
    know how Land's method could be used for color printing but it's a pity
    that no-one has tried recently. Land, of course, used two projectors,
    one for low and one for high wavelengths.

    --
    Jim Silverton (Potomac, MD)

    Extraneous "not" in Reply To.
    James Silverton, Feb 13, 2013
    #15
  16. Paul Ciszek

    Robert Coe Guest

    On Tue, 12 Feb 2013 11:12:48 +0000 (UTC), (Paul Ciszek)
    wrote:
    :
    : In article <>,
    : RichA <> wrote:
    : >
    : >Human vision extends into the violet (about 1% efficiency) and can see
    : >it. CMOS cameras are about 50% efficient in the violet and probably
    : >renders it by combining red and blue filter elements.
    :
    : Being sensetive to a wavelength of light, and being able to distinguish it
    : as a color are two different things. B&W film responds to many colors of
    : light, but doesn't differentiate them. Color film used to respond to UV,
    : but it would just show up as blue. IR response was weird.
    :
    : In order for a digital camera to render violet light as purplish, as in
    : the rainbow picture I linked to, both the red pixels and the blue pixels
    : would have to be responding to violet light. I understand why this
    : happens in the human eye--as I explained, it's because we don't have a
    : separate "red" detector as such--but in order for the bayer filter of a
    : color digital camera to render violet as purple rather than blue, the
    : "red" filters would have to pass the long wavelengths, block the green
    : and blue wavelengths, and pass again at the very shortest wavelengths.
    : This is not easy to do, especially if your filter has to be deposited
    : onto a semiconductor surface rather than mixed up in a vat by Schott
    : Glass.
    :
    : >There is no such thing as a reddish-blue colour, at least not
    : >practically, since you cannot resolve the photons of red and blue
    : >reflections so violet is what you see.
    :
    : There most certainly is such a thing as "reddish blue". Any purple you
    : see printed, rendered on a monitor, or projected on a movie screen
    : contains both red and blue light. Violet light exists in nature and
    : can be produced by some light sources, but any form of media fakes it
    : using purple--aka "reddish blue"--instead. Ironically, so do violets--
    : they are purple rather than truely violet.

    My wildass guess is that some radiation in the near-ultraviolet is perceived
    in the human eye (and maybe even more so in digital sensors) as some higher
    harmonic of red. This might not be inconsistent with the effect that others in
    this thread have reported, viz. ultraviolet flower petals being captured as
    pink.

    Bob
    Robert Coe, Feb 14, 2013
    #16
  17. Paul Ciszek

    Me Guest

    On 14/02/2013 10:39 a.m., Paul Ciszek wrote:
    > In article <kfe8b3$sh6$>,
    > Me <> wrote:
    >>>

    >> The red filter in the Bayer array probably passes some violet light.
    >> Here are some transmission charts (for photographic filters).
    >> This red filters shows a (second) "transmission peak" at the violet end
    >> of the visible spectrum. Perhaps - more correctly - the chart should be
    >> inverted so you look at it as an absorption peak rather than
    >> transmission peak.
    >>>

    >> http://www.schneiderkreuznach.com/pdf/filter/bw_filter_transmission_curves.pdf
    >> (see page 2)

    >
    > "Die von Ihnen gesuchte Seite existiert leider nicht oder der Link ist
    > veraltet. Bitte verwenden Sie die Suche oder kehren Sie zur Startseite
    > zurück, um die gewünschten Informationen zu finden."
    >

    Hmmm - it was there yesterday...
    Cached on my PC - here's a screen capture of the red filter chart:
    > http://oi46.tinypic.com/23w4mll.jpg
    Me, Feb 14, 2013
    #17
  18. Paul Ciszek

    Paul Ciszek Guest

    In article <kfhn67$irl$>,
    Me <> wrote:
    >>

    >Hmmm - it was there yesterday...
    >Cached on my PC - here's a screen capture of the red filter chart:
    > > http://oi46.tinypic.com/23w4mll.jpg


    The secondary peak at 500nm actually lies between green and blue. That
    must screw things up a bit. The response does start ramping up again
    just as the graph quits at 400nm, but that contribution is weaker than
    the blue-green peak at 500nm. I would think that the false response from
    blue-green would dominate the "purple" rendering of anything using this
    filter--Anything blue is going to be considered kind of purple. And if
    you try to correct for that, I think you will end up also eliminating the
    weaker genuine response to violet.

    --
    Please reply to: | "We establish no religion in this country, we
    pciszek at panix dot com | command no worship, we mandate no belief, nor
    Autoreply is disabled | will we ever. Church and state are, and must
    | remain, separate." --Ronald Reagan, 10/26/1984
    Paul Ciszek, Feb 14, 2013
    #18
  19. Paul Ciszek

    Me Guest

    On 15/02/2013 8:54 a.m., Paul Ciszek wrote:
    > In article <kfhn67$irl$>,
    > Me <> wrote:
    >>>

    >> Hmmm - it was there yesterday...
    >> Cached on my PC - here's a screen capture of the red filter chart:
    >>> http://oi46.tinypic.com/23w4mll.jpg

    >
    > The secondary peak at 500nm actually lies between green and blue. That
    > must screw things up a bit. The response does start ramping up again
    > just as the graph quits at 400nm, but that contribution is weaker than
    > the blue-green peak at 500nm. I would think that the false response from
    > blue-green would dominate the "purple" rendering of anything using this
    > filter--Anything blue is going to be considered kind of purple. And if
    > you try to correct for that, I think you will end up also eliminating the
    > weaker genuine response to violet.
    >

    We don't know what the absorption spectrum for the red dyestuff used in
    Bayer RGB CFA is.(these are just red colour filters) If you can find
    that out, then perhaps the answer will be obvious.
    Me, Feb 14, 2013
    #19
  20. Paul Ciszek

    Robert Coe Guest

    On Wed, 13 Feb 2013 17:29:10 -0500, James Silverton
    <> wrote:
    : On 2/13/2013 1:27 PM, Alfred Molon wrote:
    : > In article <afde856a-b41a-4960-9ebb-
    : > >, RichA says...
    : >> Violet isn't the major problem colour with sensors, green is.
    : > Why?
    : I heard Edwin Land lecture on his two color theory and was totally
    : convinced by his demonstrations. About the most spectacular was when he
    : used the two sodium D-lines (both yellow) to reconstruct color. I don't
    : know how Land's method could be used for color printing but it's a pity
    : that no-one has tried recently. Land, of course, used two projectors,
    : one for low and one for high wavelengths.

    My brother and I did a demo of the Land effect in my grandfather's dining room
    50 years ago. It was pretty convincing, but it clearly (I think) depended on
    the workings of the brain as much as on those of the eyes.

    Bob
    Robert Coe, Feb 15, 2013
    #20
    1. Advertising

Want to reply to this thread or ask your own question?

It takes just 2 minutes to sign up (and it's free!). Just click the sign up button to choose a username and then you can ask your own questions on the forum.
Similar Threads
  1. zxcvar
    Replies:
    12
    Views:
    893
    (Pete Cresswell)
    Jan 4, 2004
  2. g n p

    Minus Violet Filters

    g n p, Jun 20, 2005, in forum: Digital Photography
    Replies:
    0
    Views:
    325
    g n p
    Jun 20, 2005
  3. g n p

    Minus-Violet filters

    g n p, Jun 22, 2005, in forum: Digital Photography
    Replies:
    4
    Views:
    526
    Nostrobino
    Jun 24, 2005
  4. wagwheel

    Cameras--Cameras--Cameras

    wagwheel, Mar 31, 2007, in forum: Digital Photography
    Replies:
    4
    Views:
    422
    Mark B.
    Apr 1, 2007
  5. Blig Merk
    Replies:
    40
    Views:
    1,579
    Doug Jacobs
    Mar 3, 2008
Loading...

Share This Page