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. Yes, but that contributes colour information, not spatial information,
    so the multiple colour samples all contribute to the *same* pixel.

    There have now been at least a dozen postings pointing this out to you.
    Don't you read?
    And nobody has ever suggested that the multiple colour measurements made
    by Kodachrome at one spatial location created more than one pixel. So,
    just like Kodachrome, the X3 sensor produces one pixel from 3 layers of
    data. The Kodachrome reference demonstrates that you are wrong, so
    why include it?

    Dave
     
    Dave Martindale, Jun 15, 2004
    1. Advertisements

  2. Absolutely everyone in this discussion understands that, and has
    understood it since the Foveon cameras were introduced. So why bring it
    up?

    What everyone disagrees with is your *next* step in the argument, where
    you try to call each of the 3 sensors a pixel. Loosely, they are all
    components of one pixel. If you want to be more precise, they all
    contribute data to one pixel in the camera output image.

    Dave
     
    Dave Martindale, Jun 15, 2004
    1. Advertisements

  3. I'm starting to believe that too. He has an incredible talent for
    not following the line of an argument. He tries to explain over and
    over again facts that everyone has already understood (e.g. that there
    are 3 measurements at every photosite) while ignoring arguments about
    the real difference in positions (whether these should be counted as 1
    or 3 pixels), with a few utterly stupid semantic ideas thrown in (e.g.
    that the word megapixel might mean something independent of the word
    pixel).

    He's not providing any useful information to anyone, and he doesn't seem
    capable of learning anything from the information presented to him. He
    won't read the archives of past discussions in this group, and keeps
    referring to notes in the dpreview Sigma forum that don't support his
    position.

    He says he's a translator by profession, so he *should* be good at
    understanding arguments and presenting his own in a sensible way. So I
    think he's doing this deliberately.

    Dave
     
    Dave Martindale, Jun 15, 2004
  4. He reminds me of a child trying to circumvent one of their parents'
    rules by arguing that there's a loophole in the wording of the rule.
    No more effectively, either.

    Dave
     
    Dave Martindale, Jun 15, 2004
  5. Paul Howland

    Paul Howland Guest

    3D has a very specific meaning in science and engineering, the third
    dimension referring to the thrid spatial dimension. Now, whilst it
    would be absolutely correct for you to describe the *construction* of
    each Foveon pixel as three dimensional, it is absolutely misleading to
    describe the resulting *data* as 3D. It is not. It contains no
    information on the third dimension at all. None. To describe it as
    such is just misleading.
    This analogy is pointless. You are measuring three discrete and
    independent events here: the distance of the red car, the blue car and
    the green car. Of course you need three measurements.

    This bears no relationship to measuring the light falling on a pixel
    location. This is not a set of three discrete events (light is not
    actually composed of a mixture of red, green and blue as I'm sure you
    know) and so whatever the camera does will be some kind of
    approximation. Red, green and blue are chosen as they provide
    reasonably orthogonal measurement sets for attempting to reconstruct the
    full colour continuum, but you could equally well have one hundred
    narrower band colour filters (or four filters - eg. CMYK). Would you
    then contend that the camera was 100-dimensional? Of course not. Foveon
    uses three sensors at each pixel to gather colour information for a 2D
    image. Nothing more, nothing less. RGB is reasonably convenient as it
    matches the human eye's own colour reconstruction process, but there's
    nothing inevitable about it.

    There is nothing three dimensional about the process - unless you argue
    that because the Foveon pixel construction has some depth in the chip
    that the image is therefore 3D. In which case, I note that the Bayer
    pixels also have some kind of finite depth in the silicon too and so are
    equally three dimensional.
    A fairly pointless and misleading contention though. The fact that each
    sensor has some depth in the silicon has no relevance to the formation
    of the final image or the "3D qualities" of that image. The fact that
    the sensor also clocks out the information into the memory actually
    makes the sensor 4D, by the way. Even better.
    I think the general concensus amongst reviewers is that the SD10 has a
    resolution comparable to a 6MP Bayer sensor downsampled to 3.4MP. The
    issue is really whether the Foveon produces better colour information as
    a result of its approach. For the moment, at least, the answer to that
    is no. For good lighting the results are comparable. In poor
    lighting/high ISO's, the results of the Foveon are considerably worse,
    particularly with color shifts. This, I guess, may indeed be due to the
    3D construction of the sensors and the need for the light to penetrate
    extra depth.
     
    Paul Howland, Jun 15, 2004
  6. Paul Howland

    Paul Howland Guest

    Laurence Matson wrote:

    Incorrect analogy again Laurence. The three thermometers are making
    three discrete measurements of something that varies as a function of
    position. The light hitting your camera's sensor has a single spectrum.
    As noone's yet figured out a way to accurately capture that analogue
    light spectrum they approximate it by measuring it with three filters
    with different responses: one red, one blue, one green, and then try to
    reconstruct the colour spectrum from that. However, the fact that both
    Bayers and Foveons use these three colours to form an orthogonal
    representation of the colour content at that pixel is irrelevant. The
    actual thing being measured does not comprise three discrete values,
    unlike your thermometer analogy.
     
    Paul Howland, Jun 15, 2004
  7. Paul Howland

    Paul Howland Guest

    You're flogging a dead horse Laurence.

    Noone is disagreeing that the Foveon sensor has a three-dimensional
    construction with three colour measurement sites per pixel. People
    disagree that this technique either deserves to be called "3D" - which
    implies that depth information is actually *measured* - or that the
    technique increases the height/width resolution by a factor of three
    (ie. offers a 10MP image). Both claims are baseless marketing or
    misinformation. The best you can claim is that it provides more
    accurate colour measurements than Bayer technology. However, the
    results say otherwise.
     
    Paul Howland, Jun 15, 2004
  8. Paul Howland

    Crownfield Guest

    in a sense, yes, with a reservation:
    that 3rd dimension is not a spatial dimension,
    and it does not tell you which car is closer to the camera.
     
    Crownfield, Jun 15, 2004
  9. SNIP
    A 2-dimensional image.
    Why the patronizing tone. Afraid he won't respond? Sounds like preddy's
    craving for attention.
    In color, yes. Nobody says it doesn't.
    Spatially discrete in the image plane.
    If it doesn't have a location, it's a meaningless concept.
    You're too kind!

    SNIP
    If they are in the image/picture plane.
    They are not just Lionel's definitions.

    SNIP
    You are trolling along.

    Bart
     
    Bart van der Wolf, Jun 15, 2004
  10. Paul Howland

    Crownfield Guest

    but one point of light produces one physical datapoint in the image.

    lets make a neuveon, the advanced foveon.
    the neuveon has one sensel for each possible visible wavelength of
    light, lets say 3,000 sensels,
    all perfectly colocated in the third dimension.
    neuveon really can measure color and brightness perfectly.

    you would have ONE physical pixel for each set of 3,000 sensels.
    It might do a perfect job of rendering color,
    but only at one physical location of the image.
    3,000 sensels, one pixel.

    the next step will be the superneuveon,
    all the old foveon problems have been solved in the newveon.
    the neuveon will have 24 mp,
    each pixel made up of 3,000 sensels
    stacked and perfectly colocated.

    the image will be made up of 6,000 x 4,000 pixels x 3,000 colors.
    the only problem will be file size.

    no physical interpolation.
    no color interpolation.

    the camera that the prediot dreams of,
    and thinks he could get from sigma.
     
    Crownfield, Jun 15, 2004
  11. Paul Howland

    Crownfield Guest

    pure bull shit.
    look at the raw output file.
     
    Crownfield, Jun 15, 2004
  12. Paul Howland

    Crownfield Guest

    foveons guess statistically at the image.

    a tiny 3 mp obsolete sensor.
    and their newest deal is polaroid?

    surely you jest.
     
    Crownfield, Jun 15, 2004
  13. Water temperature is a 3D function: it varies north/south, east/west,
    and with depth. The world in general is 3D: you can measure light
    intensity reaching any point in 3D space.

    But a photograph is an abstraction of this: a lens captures an image
    from the light reflected or emitted in *one* direction (towards the
    lens), and the 3D structure of the scene is projected into a 2D image on
    the film or sensor. The third dimension is gone.

    The sensor's job is to measure that 2D image, and only
    within-image-plane differences in position matter to spatial resolution.
    That's why a pixel is defined as a position within the 2D image plane.

    Dave
     
    Dave Martindale, Jun 15, 2004
  14. (George Preddy) wrote in
    Hmmm - quantum mechanical accuracy :)



    /Roland
     
    Roland Karlsson, Jun 15, 2004
  15. Shhh ... Guido may wake up.

    He is nuts to, but not as fun as Preddy.



    /Roland
     
    Roland Karlsson, Jun 15, 2004
  16. Sorry Paul, but I have to disagree. Not that I think that you can
    get any useful 3D-info from the Foveon sensor, but a potental method
    for recording 3D-pictures would be to have a 3D-sensor. The plane
    where it is sharpest corresponds to the distance.

    But, all this is very near to nonsens. I don't even think that Foveon
    would call their sensor a 3D sensor. Or would they?

    More colors per pixel. More pixels per pixel. And more dimensions
    per pixel.


    /Roland
     
    Roland Karlsson, Jun 15, 2004
  17. Paul Howland

    JPS Guest

    In message <>,
    There is no apostrophe in the plural of Sigma.

    There is no RGB measured by them, either. Three bands of light are
    separated by the layers, but they do not have typical filter response
    curves nor are they necessarily centered on R, G, and B. Differential
    math is used to calculate RGB, with a lot of imprecision.
    --
     
    JPS, Jun 15, 2004
  18. Paul Howland

    JPS Guest

    In message <>,
    Everyone but you and StevieGeorge knows that there are 3 sensors at each
    2D location on a Foveon. They also know that the fact that they are
    spatially distinct has no relevance to image capture, as the
    displacement is in the most irrelevant axis.
    --
     
    JPS, Jun 15, 2004
  19. Paul Howland

    JPS Guest

    In message <>,
    Different wavelengths in the same *PIXEL*.
    --
     
    JPS, Jun 15, 2004
  20. Paul Howland

    Big Bill Guest

    But in such a sensor, the 3D elements (sensors) would need to collect
    data from *different* samples of light, which the X3 does not do.
    There's no way to get meaningful 3D images from the X3 because all 3
    sensors at each sensing site are in a straight line which is
    perpendicular to the focal plane; they all sample the same light,
    minus filtered parts of the spectrum.
    I seriously doubt they would make a claim of 3D images, for the above
    reason.
    Bill Funk
    Change "g" to "a"
     
    Big Bill, Jun 15, 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.