Re: 25 Reasons to Aviod the SD-10 (was 15 Reasons to Aviod the SD-10)

Discussion in 'Digital Cameras' started by Paul Howland, May 9, 2004.

  1. boy Preddy, you just keep making things up, what the **** is 1.5MP
    optical.

    You have no idea what a pixel is do you?

    GK
     
    grant kinsley, May 16, 2004
    #21
    1. Advertisements

  2. In other words you are making shit up. a pixel is one picture element.
    a Canon 10D and 300D has 6 million of them. A foveon sensor has 3.43
    million of them. No matter how you slice it one picture element is 1
    picture element. Whether it is interpolated via a mosaic (as our eye
    does it) or whether it is sampled as a faux RGB at one site and
    heavily corrected (considering the silicon wells are not RGP, nor are
    they partyicularly accurate) you can't change the spatial issue that
    the Bayer sensor has nearly twice the spatial information.
    No george you are a liar. luminance is fully sampled at every
    photosite. You really need to learn what luminance and chrominance are
    and why luminance is substantially more important in our visual
    system. You also need to learn what aliasing is all about. Just
    because you think the Foveon image looks sharp, it is only because of
    the highly inaccurate spatial mapping produced by the lack of an AA
    filter.

    No matter how many time you lie, the result is it's still a lie.

    GK
     
    grant kinsley, May 16, 2004
    #22
    1. Advertisements

  3. SNIP
    Wrong. Optical resolution is what the lens delivers to the sensor. The
    sensor samples the continuous signal, which then becomes a discrete sampled
    image after quantization of the signal.
    Nonsense statement.
    Which lowers it's resolution per pixel even further.
    You don't know how it works.

    Bart
     
    Bart van der Wolf, May 16, 2004
    #23
  4. Paul Howland

    JPS Guest

    In message <>,
    That means that the Bayer figures are not a scam, then. If you want a
    more specific description, then *you* need to qualify.
    They don't say the camera captures 6M full-RGB pixels.
    The mosaic in the Bayer cameras cut out less light than the SD9's lack
    of microlenses does. 70% of the light hitting the SD9 sensor is
    absorbed as heat, or bounces around to lower contrast or cause
    artifacts.
    I knew about Bayer CFAs after I bought my first digital camera 5 years
    ago. I had known about it for 4 years before buying my 10D. I was
    excited when I heard about Foveon X3 technology. I was totally
    unimpressed when I saw the output of a real Sigma DSLR.
    Full something-or-other, but not RGB.
    Foveon/Sigma is the scam; claiming better color but giving worst color.
    --
     
    JPS, May 16, 2004
    #24
  5. SNIP
    If you call a Bayer CFA sensor "Monochrome", then the Sigma is 3-color, not
    full-color. It is also far from RGB because of the poor color separation, so
    it will produce extra noise in postprocessing.

    Bart
     
    Bart van der Wolf, May 16, 2004
    #25
  6. in video and digital photo applications a pixel is by definition a
    single picture element (pixel is short for PICture ELement)
    There are 6 million pixels, I don't give a shit how you like to
    subdivide, there are 6 million elements, accurate color is reproduced
    at all of them due to very accurate interpolation by the software.
    Bluntly, the mathematics involved are far beyond your understanding.
    But suffice it to say, it is closely modeled on the fact that our eyes
    are much more sensitive to luminance than chrominance (even video uses
    that aspect, it is why a component signal uses a full 3rd of the
    signal for luminace only).
    You're jealous because you can't afford more than a computer webcam.
    each one samples 3 times, not using an RGB definition, but a highly
    variable 3 well definition based on the natural filtering abilities of
    silicon, unfortunatly significant discrepancies occur due to variable
    voltaics that occur between the wells and from voltaic differences
    between adjacent sensors.

    I suggest you look up an issue called fixed pixel noise and you may
    begin to understand why the three well sensing system is inherently
    inferior to interpolation (fixed pixel noise will be multiplied by
    three when there are 3 sensors at one site).

    Sigma disadvantage: the sensor system of using silicons natural filter
    is not precise in it's colour definition, it uses a gradient over
    which to sense those colours, the consequence is it is not RGB, and
    suffers form high degrees of desaturation for accuracy (as previous
    examples of unprocessed Sigma Raw files have shown). The software to
    then interpolate the true colours, unfortunatly as good as it is ,
    suffers from the issue of having to deal with the imperfect filtering
    of colours through the depth of the silicon, consequences include
    unnatural colours (yellow skin, purple skies) that are uneven in
    presentation

    lack of an AA filter causes luminance to snap to the centre of the
    pixel. each pixel is, therefore, by definition, not truly
    representative of the actual luminance, although giving a sense of
    sharpness, this is false, as any diagonal line or repeating pattern
    well shows on the Sigma. As well this often leads to artifacts in
    textures. certain fine textures such as tree bark and striated stone
    lose their natural look as the lack of an AA filter causes the detail
    to snap to a grid. the lack of an AA filter simply makes it quite
    impossible for fine details to fall where they are actually supposed
    to fall. They are moved to the nearest pixel centre by default.

    No matter how you slice it the Sigma only has 3.43 million spatial
    locations, no matter how much info is in those locations is
    irrelevant. One does not produce an accurate representaion of
    something by samling 343,000 colours at 10 locations. And that is the
    whole basis of your argument.

    GK
     
    grant kinsley, May 16, 2004
    #26
  7. Paul Howland

    Ray Fischer Guest

    That's a lie, of course. Generating random numbers is not detecting
    colors.
     
    Ray Fischer, May 17, 2004
    #27
  8. You apparently are confused about the distinction between sensing and
    output.
    Yes, closer but still wrong. It 'senses' roughly one third of visible
    spectrum. That is quantized into 12-bits (=4096 shades) before gamma
    adjustment.

    Bart
     
    Bart van der Wolf, May 17, 2004
    #28
  9. Paul Howland

    JPS Guest

    In message <>,
    No, that's not what I thought when I bought the 10D. I bought the 10D
    after already knowing the difference between a CFA and a Foveon sensor.
    I saw output from both systems, and chose the 10D. Guido V. posted
    links to SD9 images, and I thought the color was a bit artificial
    looking, and fine, thin objects looked like tetris pieces.
    --
     
    JPS, May 17, 2004
    #29
  10. (Georgette Preddy) wrote in
    If you have a 12 bit A/D, then the Bayer senses 4000
    levels at each site minus noise. The Foveon then senses
    3x4000 levels at each site, minus noise.

    Discrete colors? Over a large enough area, the Bayer senses
    those 64 billion colors minus noise, same as the Foveon.
    But the Foveon noise is larger and the Foveon color sensitivity
    is strange, so actually, it is worse.

    Per pixel color accuracy might be better for Foveon. But that is
    not all that important as the eye does not have all that good
    spatial resoultion for color detection.


    /Roland
     
    Roland Karlsson, May 17, 2004
    #30
  11. Paul Howland

    JPS Guest

    In message <[email protected]>,
    How could it be better, when smooth gradients have distinct blotches in
    the blue and green channels? The only thing Foveon seems to be better
    at is resolving blue vs green vs red at the nyquist. The quality of
    those blues greens and reds, however, may be way off.

    --
     
    JPS, May 18, 2004
    #31
  12. Paul Howland

    Crownfield Guest

    Quantity, but not quality.
    funny.
    foveon can not reproduce any color consistently or accurately.
     
    Crownfield, May 18, 2004
    #32
  13. Paul Howland

    Crownfield Guest

    but they just can not output it precisely.
     
    Crownfield, May 18, 2004
    #33
  14. Paul Howland

    Lionel Guest

    Kibo informs me that (Georgette Preddy)
    stated that:
    This is a lie, & every Foveon picture you link to proves it.

    Where are those photos you claim to have sold, liar?
     
    Lionel, May 18, 2004
    #34
  15. I think you got that mixed up. Should of course be:

    "Foveon color is perfectly virtual"

    -JP
     
    Jukka-Pekka Suominen, May 18, 2004
    #35
  16. Inventing numbers again, are we? Now, if you could explain the
    following, please:

    "JPEG exclusively which drops the palette to only 1 of 256 possible
    shades of color information per output pixel"

    Are you claiming that an 8-bit JPEG can only contain 256 colors, or what
    are you trying to say. Your point eludes me on this one. And where did
    you pull that 68 billion figure from?
    The fact is that every digital sensor discretely samples information
    with a resolution based on the A/D converter. Only analog devices can
    have any chance of actually measuring anything.

    -JP
     
    Jukka-Pekka Suominen, May 18, 2004
    #36
  17. Paul Howland

    Mick Sterbs Guest

    Care to name your source for that, your latest lie?
    You're just bursting with stuff you make up and cannot attribute to any
    source whatsoever.
    You are a waste of oxygen.
     
    Mick Sterbs, May 18, 2004
    #37
  18. That's utter nonsense. No 3-colour camera has any chance of even
    approximately measuring wavelength of a single-frequency light source,
    or the spectrum of an arbitrary source. For that, you need a
    spectroradiometer, with about 32 wavelength bands spaced 10 nm apart or
    64 wavelength bands spaced 5 nm apart. (That's 32/64 samples for every
    pixel!).

    With a 3-colour camera, an arbitrary incoming image is converted to
    just 3 numbers. Almost all spectral information is lost. But what you
    *can* do is measure in a colour space that matches human colour
    vision. If you do that, you can recreate a colour which looks the same
    to the human eye, even if it has a very different spectrum from the
    original light (and even though the two colours probably do not look
    the same to a cat or other animal).

    Bayer cameras do try to measure colour in this way. Unfortunately, the
    Foveon sensor's raw colour response is nowhere close to the human eye
    colour matching functions. Foveon has published the spectral response
    of a filter that could be added to a Foveon sensor to mostly correct
    this, but there's no evidence the Sigma cameras actually use it.

    Yet another example of "George" writing something that sounds scientific
    and precise when in fact it's complete nonsense.

    Dave
     
    Dave Martindale, May 18, 2004
    #38
  19. Perhaps the problem with the X3 is that it's using a CMOS process that it
    Or it may be that selective absorption of photons as a function of
    energy (wavelength) and silicon depth is fundamentally a poor way of
    measuring colour, and can never be made to work well.

    Dave
     
    Dave Martindale, May 18, 2004
    #39
  20. I said might :)

    The actual results from Sigma cameras tell another story,
    just as you say.

    But (and now Goerge(tte) will of course use this totally faulty
    to pretend I admit that Foveon is better :) each Bayer pixel does
    interpolate color. Therefore, the color of each pixel might be wrong.
    Note that I said "might" again :).

    The lack of anti alias filter on the Sigma sensors makes both
    luminance and color wrong of course. And the noise that comes from
    converting from the strange color space also makes the color
    accuracy low.

    So - it is a very strong might :)


    /Roland
     
    Roland Karlsson, May 18, 2004
    #40
    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.