JPEG 2000: which camera will support it? When?

Discussion in 'Digital Photography' started by John Faughnan, Jan 28, 2004.

  1. []
    Maybe, but the Nikons I use seem to do quite a good job (for the type of
    shots I take). Most of my presentation is individual images on a CRT or
    LCD display rather than prints where you might view more than one image at
    a time.

    []
    Let's concentrate on what the majority use and enjoy!

    Cheers,
    David
     
    David J Taylor, Feb 3, 2004
    #41
    1. Advertisements

  2. Guido,

    while reading messages in a related thread, one question made me
    curious. Do the pixel color filters in a Bayer sensor filter out
    almost all off-color light or do they only reduce the other
    colors somewhat but do not block them entirely?

    I hope I formulated the question such that it can be understood.
    In other words, does a green filtered pixel only measure green
    light or does it only have a somewhat pronounced green
    sensitivity but still a lot of general luminance information?

    By the way, I think that these religious Foveon-Bayer wars tend
    to miss the point a little. As long as the sensors comprise
    single pixel sensors, the question is, what is the optimal color
    filter distribution? And perhaps whether the pixels should be
    arranged in squares or like bee-hives. I also wonder whether any
    manufacturer has got the idea to have four pixels filtered not
    as red-green-green-blue, but as white-red-green-blue for better
    luminance information.

    Hans-Georg
     
    Hans-Georg Michna, Feb 4, 2004
    #42
    1. Advertisements

  3. Color perception is a somewhat difficult science and sometimes
    controversal matter.
    According to tri-stimulus color theory we make up all colors
    from only 3 basic (primary) colors. The primary colors have a
    certain response function over the wavelenghts spectrum which
    just peaks at a Red, Green, or Blue point. There are compromises
    with advantages and disadvantages whether to make the response
    curves more narrow or more wide. So it is not the full truth
    if people complain about possibly wide response curves in the
    Foveon sensor which would lead to bad color separation. A
    certain wideness is necessary for appropriate output color gamut.
    See the figures in slides nr. 33-37 in the Foveon presentation
    to get an idea:

    "Silicon as a Color Filter"
    "Silicon Color Separation"
    "X3 Spectral Response Curves"
    "Human Cone Spectral Response Curves"
    "Color-Matching Functions"

    http://www-inst.eecs.berkeley.edu/~cs39j/fa03/session03a.html
    The mosaic filter approach is an unsuccessful approach, whatever
    you do. You must avoid that. (Your white element wouldn't help
    due to poor resolution.) Foveon seems to be a succcessful
    solution - another approach was taken in video with 3-chip design
    and color-separating prism, but with high cost and other problems.

    Regards
    Guido
     
    Guido Vollbeding, Feb 4, 2004
    #43
  4. John Faughnan

    Mark Herring Guest

    The idea is to match the response of the color filters in the eye. If
    there is too much out of band response, there will be errors.

    **************************
    Mark Herring, Pasadena, Calif.
    Private e-mail: Just say no to "No".
     
    Mark Herring, Feb 4, 2004
    #44
  5. John Faughnan

    eawckyegcy Guest

    I asked you before: are you this stupid in real life, or is this just
    some sort of act you are putting on?
    The CFA is evidence of what is called "engineering". While you
    blather on metaphysically about "unsuccessful approaches" and things
    we must "avoid" (Lord Kelvin (1895): "Heavier than air flying
    machiners are impossible"), it nevertheless remains a fact that CFA
    cameras produce completely acceptable (if not excellent) images
    anyways.

    "If physical reality says you are an ass, you are an ass." (Uncle Al)
    Success is defined by something that currently produces _worse_ colour
    than the so-called unsuccessful design?

    Your thought processes are, well, difficult to discern.
    Still cameras have a splendid advantage over video: they can take a
    few seconds to digest the data and produce a better image. When you
    are given only a 1/30th of a second you have to use a different
    solution or accept the consequences.

    But thats more of that over-your-head engineering for you.
    Quality/cost tradeoffs are made all the time in the real world: deal
    with it.
     
    eawckyegcy, Feb 4, 2004
    #45
  6. John Faughnan

    eawckyegcy Guest

    See http://astrosurf.com/buil/us/digit/spectra.htm for measurements of
    the Canon 10D's response. I would expect most cameras to be more or
    less the same.
    Note there is luminance information in all the channels of a CFA, not
    just the green. The green is "assumed" to be the luminance simply
    because there is more of it on the sensor (and, of course, because our
    eyes are more sensitive to it).

    However, I highly recommend perusing some demosaic papers because alot
    of the blather in this newsgroup re: "luminance" is mainly a
    red-herring: much, MUCH, more important is channel-to-channel
    correlation. Without this the CFA approach would indeed be a poor
    one. I've posted references in the past in this newsgroup.
    The Bayer-Foveon wars exist so we can laugh at idiots like George
    Preddy and his company of sock puppets.
     
    eawckyegcy, Feb 4, 2004
    #46
  7. John Faughnan

    Chris Brown Guest

    That one always amused me. Either he never noticed birds, or must have
    decided it was impossible to engineer a power source with a sufficient power
    to weight ratio to power something, say, the size of a swan.

    Given that swans do a pretty good job of turning fish and bread into heavier
    than air flight, he must have realised that he was on dodgy ground.

    Oh yeah, and as for Guido, I don't think anyone really takes him seriously.
    The large numbers of professional and amateur photographers getting
    perfectly good results from mosaic-sensor digital cameras every day, many of
    which are now rather superior to traditional wet-process smll format
    photography, lowers his position to that of "ill founded rant".
    I congratulate you on your tact under fire. :)
     
    Chris Brown, Feb 4, 2004
    #47
  8. Thank you, all, for the replies, but I guess I have to try to
    reformulate the question, as all the answers so far didn't
    answer my (single-bit) question.

    Let me ask this way: Does the red filter (on the respective
    pixels) in a Bayer sensor let only red light pass? Or is it a
    pink filter that lets all colors pass, only more red than the
    others?

    Hans-Georg
     
    Hans-Georg Michna, Feb 5, 2004
    #48
  9. [snip]

    The quality to compression ratio of digicams could be improved
    considerably by using dynamically generated Huffman dictionaries. This
    is the "Baseline optimized" option in Photoshop that makes JPEGs smaller
    by improving the final lossless stage of compression.

    The problem is, creating a custom Huffman table requires another pass
    over the JPEG data. The JPEG data can no longer be streamed through an
    encoder chip that has the baseline Huffman table in ROM. It needs to be
    buffered in lots of RAM for two pass processing. That wasn't practical
    in the past but it should be standard soon.
     
    Kevin McMurtrie, Feb 5, 2004
    #49
  10. John Faughnan

    Martin Brown Guest

    They are meant to do this in as far as it is physically possible.
    Usually the red filter is the only one with a really clean cutoff that
    does what it says on the can. The green and blue filters may have some
    feeble transmission in red or near IR. Easy enough to check - put a
    Wratten 25 or better Wratten 29 filter in front of your camera and shoot
    through that.

    Bayer filters are meant to be nominally pure RGB colours (unless they
    are the alternative more sensitive secondary pattern CMYG).

    My old Kodak DC-120 used to go haywire shooting solar prominences in
    monochromatic red hydrogen light at 656nm - all channels saw signal.

    Regards,
     
    Martin Brown, Feb 5, 2004
    #50
  11. I don't think it will be practical soon, but there's a better alternative.
    Wait another 5 or so years, then they can apply the JPEG arithmetic coding
    option (the patents expire then).
    Arithmetic coding needs only a single processing pass, and it compresses
    even better than the optimized Huffman option! You can try this today
    with the Jpegcrop program, see http://jpegclub.org/jpegcrop/ for more
    details.

    BTW, JPEG2000 uses arithmetic coding by default, that's one reason why
    industry doesn't adopt it, due to patent license requirements.
    But J2K is inferior and obsolete anyway, so not a big deal there.

    Regards
    Guido
     
    Guido Vollbeding, Feb 5, 2004
    #51
  12. I don't expect to be taken seriously by incompetent people.
    I'm too long in this business to be impressed by the fakes
    and hoaxes from so-called 'experts', or by the acceptance
    of crap by the crowd. I only hope that they don't bother
    me too much with their junk.

    Regards
    Guido
     
    Guido Vollbeding, Feb 5, 2004
    #52
  13. John Faughnan

    Lionel Guest

    You & Geroge aren't taken seriously by /anyone/, competant or or
    otherwise.
     
    Lionel, Feb 5, 2004
    #53
  14. John Faughnan

    imbsysop Guest

    wow .. and you were carved right out of the thighs of Zeus I presume ?
     
    imbsysop, Feb 5, 2004
    #54
  15. John Faughnan

    Azzz1588 Guest


    Have no fear, no one here ever takes you seriously............




























    "Only a Gentleman can insult me, and a true Gentleman never will..."
     
    Azzz1588, Feb 5, 2004
    #55
  16. RAW from a Bayer camera is far away from a natural photograph,
    The human eye is missing 99,9999 % of the natural color spectrum (which is
    continuos and not restricted to 3 "basic" colors) !

    -Michael
     
    Michael Schnell, Feb 5, 2004
    #56
  17. Turning it round, then, considering that the majority of cameras today
    Maybe it really does. If a camera (like the human eye) offers more
    resolution on grayscale than on color, the compression could be more tight
    on color than on grayscale. Or, regarding a Bayer sensor, tighter on red and
    blue than on green.

    AFAIK, JP2000 provides a color space and multiple differently compressed
    monochrome pictures. So an optimum color space for the appropriate sensor
    can be selected.

    -Michael
     
    Michael Schnell, Feb 5, 2004
    #57
  18. John Faughnan

    Lionel Guest

    Kibo informs me that Michael Schnell
    Which is exactly what JPEG compression does, for exactly the same reason
    that mosaic sensors distribute colours the way they do. In other words,
    they both take advantage of the way the eye works.
     
    Lionel, Feb 5, 2004
    #58
  19. John Faughnan

    Alfred Molon Guest

    Speak for yourself.
     
    Alfred Molon, Feb 5, 2004
    #59
  20. John Faughnan

    jpc Guest

    If you want to see how much of each color the Bayer filters on your
    camera passes or blocks do the following.


    Download the tif image of visible spectrum you find at
    http://home.earthlink.net/~skywise711/LasersOptics/Reference/reference.html.
    It is a reasonablely intense image with wavelength markers

    It is also a long narrow image so I cropped and resampled it to print
    it on a 4 by 6. You could also crop out the violet section beyond the
    dark blue since that is only an artistic attempt at showing UV.

    Print it out and take an image under diffuse sunlight.

    Bring the image into a photo editor where you can display the
    individual channels; red, green and blue. I used Photoshop but I
    believe you could do this with any medium to high end image editor.

    Duplicate each channel to make a new image and convert to greyscale.
    Where it's bright you've collected colors, where it's black you've
    blocked them.

    Tile your images and compare. If you need numbers, download ImageJ so
    you can plot out the relative intensities of each grey scale image.

    With my camera the red channel completely blocked all blue but passed
    yellow and some green, The green channel blocked the red, passed the
    yellow and green plus more blue than I expected. The blue channel
    passed the blue and after dropping quickly to about 10 percent slowly
    fell without completely blocking either the red or green.

    jpc
     
    jpc, Feb 5, 2004
    #60
    1. Advertisements

Ask a Question

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

You'll need to choose a username for the site, which only take a couple of moments (here). After that, you can post your question and our members will help you out.