New Polaroid/Foveon Digicam

Discussion in 'Digital Photography' started by R2D2, Feb 9, 2004.

  1. If you don't measure the spectrum of light with a response that matches
    the luminance sensitivity of the eye (or RGB filters that can be
    combined to produce this response), you can only guess at luminance.

    Bayer misses, because it only measures one colour at each pixel. But
    unless colour is changing very rapidly, the Bayer output is very close
    to being correct.

    Foveon misses, because the "RGB" filter responses are nothing close to
    what they should be to capture proper luminance. Judging from example
    photos, it's often wrong.

    So there are no consumer cameras that accurately measure luminance under
    all conditions, Foveon included.

    Since we can't have perfection, we might as well pick whatever looks
    best.

    Dave
     
    Dave Martindale, Feb 11, 2004
    1. Advertisements

  2. Dave Martindale, Feb 11, 2004
    1. Advertisements

  3. R2D2

    Nils Rostedt Guest

    According to dpreview, the SD9 outputs:

    "Image sizes . 2268 x 1512 (HI)
    . 1512 x 1008 (MED)
    . 1134 x 756 (LOW)
    * Can output a double-size 4536 x 3024 from Photo Pro software "

    Since when is the PC+Windows that is required for upsampled 4536x3024
    included in the price of the SD9/10, George?
     
    Nils Rostedt, Feb 11, 2004
  4. R2D2

    JPS Guest

    In message <>,
    You are not making sense. The SDx cameras have far more false
    modulation than the 6MP Bayer DSLRs. The Bayer cameras are
    frequency-limited to slightly below the nyquist for luminance, and half
    the nyquist for chrominance. Frequencies above these don't record, and
    frequencies below these are recorded very accurately.
    --
     
    JPS, Feb 11, 2004
  5. The sensor has 10.29M photosites, each with their own location, but the
    images have 13.72M pixels using Bayer's level of interpolation.
     
    George Preddy, Feb 11, 2004
  6. A PC+Windows is not required. But you can use a PC+Windows to take the
    camera's 10.3MP RAW files and output it at 3 different resolutions. In the
    SD9's "Medium" resoution mode, the RAW software outputs 6MP-interpolated
    images, using the same degree of interpolation as 6MP-interpolated Bayer
    images.
     
    George Preddy, Feb 11, 2004
  7. No - it don't. You cannot just count the sensor elements
    when you compare sensor technologies. The new Fuiji sensors
    have twice as many sensor elements. This will not double
    the resolution, it will (hopefully) increase the dynamic
    range.

    Guido says that Bayer sensors does not record any correct
    RGB value anywhere - therefore it does not record photographs
    (or was it images?).

    You say that Bayer record images in RGGB squares, therefore
    you get an image with half the resolution that normally is claimed.

    You are both wrong - and you don't understand how Bayer images
    are constructed. Personally, I would rather have pure RGB values
    at each point - but currently Bayer is the next best - and
    Foveon the third best.


    /Roland
     
    Roland Karlsson, Feb 11, 2004
  8. It isn't possible to "discredit" simple color resolution pictures. The 10D
    has about 1/3rd the color resolution of the SD9, it is a fact. The 1Ds is a
    little worse than the 10D, blurry but not as blurry, and has some extra FOV
    to show for it.
     
    George Preddy, Feb 11, 2004
  9. The eye has nothing to do with anything. Luminance must be guessed unless
    it is measured, Bayer has to guess.
     
    George Preddy, Feb 11, 2004
  10. Bart van der Wolf, Feb 11, 2004
  11. Oops - you got it nearly right there George. This is
    actually what measurements show.



    /Roland
     
    Roland Karlsson, Feb 11, 2004
  12. R2D2

    Jürgen Eidt Guest

    I would rather have CIELAB and a monitor capable of displaying this.
    The colorspace around the RGB is rather limited.
     
    Jürgen Eidt, Feb 11, 2004
  13. Now - let us se - luminance - how is it defined?

    http://roland.lerc.nasa.gov/~dglover/dictionary/l.html

    luminance:
    In photometry, a measure of the intrinsic *luminous intensity* emitted
    by source in a given direction.

    luminous intensity:
    *Luminous energy* per unit time per unit solid angle.

    luminous energy (symbol Q):
    The energy of visible radiation, weighted in accordance with
    the wavelength dependence of the response of the human eye.

    Hmmm ... it sure looks like the eye has something to do with luminance.


    /Roland (not the same roland as roland.lerc.nasa.gov :)
     
    Roland Karlsson, Feb 11, 2004
  14. R2D2

    Lionel Guest

    Exactly as I predicted, you snipped & ignored my entire post. What a
    surprise.

    Tell me something, George, why did you buy a camera that you don't
    actually take photos with?
     
    Lionel, Feb 11, 2004
  15. :)

    Now ... CIEXYZ is assumed to be a good choice for color
    space compared to human seing. Unfortunately CIEXYZ has
    a rather strange choice of wavelengths for prime colors,
    so does the eye :)


    /Roland
     
    Roland Karlsson, Feb 11, 2004
  16. He cant do that you know. He is sitting starring at his news
    reader all hours while not sleeping. Nothing interesting to
    take any photos of there.


    /Roland
     
    Roland Karlsson, Feb 11, 2004
  17. R2D2

    Jürgen Eidt Guest

    Yes, the spectral curves overlap nicely.
    :)
     
    Jürgen Eidt, Feb 11, 2004
  18. R2D2

    Lionel Guest

    Kibo informs me that stated that:
    Neither Steve/George or Guido actually know what any of these terms
    mean.
     
    Lionel, Feb 11, 2004
  19. R2D2

    JPS Guest

    In message <>,
    The color of rendered pixels on 6 MP bayer DSLRs is far more accurate
    than the chromatic noise from a Foveon.
    --
     
    JPS, Feb 11, 2004
  20. Wrong, the sensor has 2268 columns x 1512, anything more is fraud by
    software; there can only be information at the intersections of the 2268
    columns x 1512; you are a liar.
     
    Ian Robert Walker, Feb 12, 2004
    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.