Bayer with NO anti-aliasing (Kodak Pro 14n)

Discussion in 'Digital Photography' started by K2, Nov 28, 2003.

  1. K2

    JPS Guest

    In message <>,
    Please stop referring to concentric RGB sensors as "Foveon" or "The
    Foveon idea". Millions of people had the idea before Foveon; in fact
    most people who know that color graphics use formats with separate RGB
    would naturally assume that all three are sensed for each pixel, unless
    they heard of the bayer CFA. Foveon is merely a company that promised
    to make the basic universal idea a technological reality, but have
    failed to do so with any level of quality in terms of providing data
    from which color can be easily or accurately interpolated.
    JPS, Dec 4, 2003
    1. Advertisements

  2. K2

    JPS Guest

    In message <bqh0ns$k4r$>,
    Nonsense, liar. Nobody assumed that Foveon had the same sensor pattern
    as a Bayer. Some may have called the full RGB triplet a single sensor
    (as in, "an RGB sensor", but no one assumed anything less than three
    colors being detected at each coordinate. Everyone who knew of Foveon
    knew it as something different because it did have RGB at each
    No, this whole thing started because you came in here with insane
    claims, just like you used to do with the Amiga in another newsgroup.
    It's not efficient at all, that's the problem. The overhead involved in
    trying to get a full-RGB concentric sensor makes high resolutions a
    great technologial challenge, and it really isn't worth it with current
    technology, because for the same cost you can make a bayer with less
    channel sensors than a Foveon that captures more real detail unless the
    subject is an unlikely high-frequency big change of hue with little or
    no change in luminance. Sigma decided to make up for the lack of
    resolution by omitting not only an anti-alias filter, but even
    microlenses, making the images appear "sharp" with artifacts when sharp
    lenses are used.

    This false SD9 sharpness at the pixel level is your rock of hope, as the
    SD9 images at 1:1 pixel viewing on a monitor have more "zing", but
    unfortunately, this zing is not from real detail, but from artifacts.
    Yes, high-frequency details below the nyquist are more pronounced with
    the SD9, but they are also polluted with detail from above the nyquist,
    which you can not filter out without filtering out real detail as well.
    JPS, Dec 4, 2003
    1. Advertisements

  3. K2

    JPS Guest

    In message <bqknsl$p3o$>,
    No, I think what you are looking for is "The SD9 takes 10.3M samples in
    3.43M unique spatial sensor coordinates".
    JPS, Dec 4, 2003
  4. K2

    JPS Guest

    In message <3fccf26e$0$204$>,
    Please don't call concentric RGB triplets "the Foveon concept", or "the
    foveon idea', or "foveon". They didn't invent that idea; it is a basic
    intuition. _Their_ idea was to make this basic intuition a
    technological reality, and they have failed to deliver quality goods.
    They have not found a way to filter light so that it charges the
    photosites in the correct layer.
    JPS, Dec 4, 2003
  5. K2

    JPS Guest

    In message <>,
    Obviously, there is no need to filter color aliasing with a Foveon, so
    the anti-aliasing filter can be weaker. The Sigma SD9, however, not
    only has no anti-aliasing filter at all, but doesn't even have
    microlenses to direct the light that would hit the surface of the sensor
    between photosites into the photosites.

    The result is that the SD9 is a mean, lean aliasing machine.
    JPS, Dec 4, 2003
  6. George Preddy, Dec 4, 2003
  7. There are loads of full color sensor cameras on the market, only a few are
    made by Foveon. The advantages are well understaood by all in the know, and
    they are considered utterly dominant designs compared to thier single chip
    counterparts. Foveon may have pioneered the 3 chip camera, but today there
    are many. The Foveon Pro 10M sensor is only revolutionary in that it brings
    this universally understood (in photography circles, not here) extraordinary
    advantage to a DSLR, by fabbing all 3 chips (now layers) on a a single die.
    George Preddy, Dec 4, 2003
  8. It looks noise free to me. What are you talking about?
    George Preddy, Dec 4, 2003
  9. Same for all sensors (individual) including Foveon's. Just look at the
    channel breakdown of any image. So bottom line, it doesn't matter since it
    is the same for both and Foveon will get (10.3/6.0)X as much "extra"
    luminance as a 6M sensor Bayer, any way you figure.
    George Preddy, Dec 4, 2003
  10. It does. It generates full previews that can be zoomed-in to the pixel
    level, unlike the unusable S2 Pro in RAW mode which only shows a tiny
    preview image. The SD9 doesn't store images as JPEGs because the JPEG
    format itself causes too much harm to a single-pixel-resolution image (it
    does the same to Bayer images, but they are too blurry to care with a
    4-pixel minimum resolution). And TIF is too big.
    Cropping is not magnification, just like digital zoom isn't optical. The
    test is perfect.
    George Preddy, Dec 4, 2003
  11. Just like every Bayer DSLR user here originally thought they actually had
    more sensors in their very low sensor count cameras, than in the 10.3MP SD9.
    Many here generuinely thought the vastly superior SD9 had lower resolution
    than a truly puny 3MP Bayer, not having seen this, I guess...
    George Preddy, Dec 4, 2003
  12. K2

    Mxsmanic Guest

    The image is a mass of noise. Check the channels. Each channel has a
    different pattern of intensities, even in areas that should be smooth.
    It's really bad, probably one of the worst digital images I've seen.
    Note also the serious smearing of colors in areas like the trees. Unless
    you shot that picture in a snowstorm, it's a mess.
    Mxsmanic, Dec 4, 2003
  13. Agreed, but using the silicon absorption characteristics to separate the
    spectral bands in a single chip is IMHO a new concept, i.e. the
    implementation of it. The absorption characteristics were of course know
    already, like e.g. the better penetration by long wavelengths such as IR.

    RGB triplets from single shot cameras on the other hand have already been in
    use for a long time, nothing to do with the particular sensor being
    discussed here.

    Bart van der Wolf, Dec 4, 2003
  14. K2

    Don F Guest

    Don F, Dec 4, 2003
  15. K2

    Azzz1588 Guest

    No it wont !!

    The sd foevon sensor is only 3.34 mp, the cannon 6 mp.
    So you do get 6 mp worth from the cannon, but only 3.34 mp
    worth of luminance date from the sd foevon.

    "Only a Gentleman can insult me, and a true Gentleman never will..."
    Azzz1588, Dec 4, 2003
  16. K2

    Chris Hoopes Guest

    Sounds like they are saying each discrete/independent position can be
    used without any of the others. That's fine if you want a single 3.4MP
    image, such as a very good black & white image. You just need data from all
    THREE colors to make up ONE co-lated sample. In essence, it takes 10.3M of
    data to make up a 3.4M RGB array.
    Chris Hoopes, Dec 4, 2003
  17. K2

    JPS Guest

    In message <bqn4ht$e0g$>,
    No, what they thought was that a 6mp bayer sampled from more locations
    than a 3.43mp Foveon, and they were right.
    The SD9 has 10.2M sensors. It only has 3.43M pixels.
    They all look horrible. Every one of them. This test is out of the
    scope of < 24*36mm sensors. Yours has more contrast because your penny
    has more contrast, and because your camera aliases. If I did not know
    that Lincoln pennies started in 1959 and had dates on them, I wouldn't
    have any idea what the date on your SD9 sample was (not even that it was
    a date).
    JPS, Dec 4, 2003
  18. K2

    JPS Guest

    In message <bqn3ms$dn2$>,
    The theoretical advantages. The actual chip is not working out, and it
    is used (or rather, abused) in a Sigma cameras that try to cheat the
    laws of sampling, and capture aliasing artifacts.
    JPS, Dec 4, 2003
  19. power cells have been using the 'double well' (gathering the charges at two
    different levels in the silicon) technique to increase the power generation
    for a while. There have also been sensors used to measure the wave length
    of light, though they have never been put together in an imaging area before
    Foveon. (Well to my knowledge, which admittedly is not vey extensive.)
    When I did a search on 'silicon absorption' I turned up quite a few devices
    that exploited the fact that the absortion lenth in silicon is frequency
    Gherry Bender, Dec 5, 2003
  20. K2

    JB Guest

    Furthermore, it doesn't stop all frequencies above that zero. The
    response has zeroes at all multiples of the sampling frequency but
    (admittedly diminishing) finite transfer at intervening frequencies.
    E.g. 3/2 * the sampling frequency will be attenuated somewhat, but
    still passed.

    JB, Dec 5, 2003
    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.