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

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

  1. K2

    Rafe B. Guest


    One more time: jaggies have nothing to do with aliasing.

    The monitor you use to view your image has no impact
    on the presence or absence of aliasing.

    Aliasing happens as the image is being formed, at the
    instant of the exposure. It can't be undone.


    rafe b.
    http://www.terrapinphoto.com
     
    Rafe B., Dec 2, 2003
    1. Advertisements

  2. K2

    JPS Guest

    In message <>,
    BZZZZZZZZZZZZZZT - This doesn't happen to other image-sources when they
    are over-sharpened. Next theory.
    --
     
    JPS, Dec 2, 2003
    1. Advertisements

  3. K2

    JPS Guest

    In message <>,
    You haven't seen a Foveon with a proper AA filter, have you?

    Have you even conducted my experiment blurring and pixellating the
    chroma channels only, at 100% zoom, with the best Foveon image you can
    find? I'm not asking you to post anything; I am just asking you to try
    it, and comment. Until you can do that, and give your explanation, you
    will be stuck in Color Kindegarten.
    So do any shortcomings Bayer images have; you keep ignoring *THAT*.

    --
     
    JPS, Dec 2, 2003
  4. K2

    JPS Guest

    In message <>,
    This image is optically soft, and contains chromatic abberation, which
    is a blur of sorts, as any potential for that over-sharp look is spread
    based on hue. For an example, look at the dark spire on the left. One
    side of it has green-blue edges, and the other has magenta-red edges.

    Looks like a wide-angle lens, wide open. The sharpest lenses are
    telephotos, generally speaking; the shorter the focal length, the more
    difficult it is to design a sharp lens, especially zooms.

    Show me a *SHARP* SD9 image without artifacts - you can't, because there
    is no such thing. Optical sharpness *GUARANTEES* visible artifacts,
    unless you are taking sharp pictures of a solid blue sky.
    --
     
    JPS, Dec 2, 2003
  5. K2

    JPS Guest

    In message <>,
    You don't know what you're talking about; the effect of the green rope
    is only possible without anti-aliasing; sharpening in software alone can
    not do that to a properly filtered sensor. Oversharpening scanned
    images, or Bayer images downsampled to 25%, will never give results like
    this. It is only possible by omitting the AA filter.

    You're proving to be just as mental as SteveGeorge; as soon as you start
    to see the light of reason, you jump back to your cultist perch. I had
    a feeling when you first started to sound open-minded and reasonable,
    that it wasn't going to last for long. Maybe the others are right about
    you being SteveGeorge. I doubted it at first, but this is starting to
    smell like a Socratic Puppet Show.


    --
     
    JPS, Dec 2, 2003
  6. K2

    JPS Guest

    In message <bqfit5$hqo$>,
    It speaks nothing at all. I don't even have any shots anywhere to post
    a link to, and I'm certainly not going to start up a web site just for a
    slimy character like you. I already know what 10D and SD9 images are
    like, and so do most of the people here. The 10D images are generally
    softer, because they have an antialiasing filter, and the default for
    the converters is very little sharpening. The SD9 images look very
    sharp, if the lens is sharp, but not with the real detail of the scene,
    but rather, with artifacts, due to a lack of an anti-aliasing filter and
    microlenses. The SD9 can detect hue changes where there is no luminance
    change at a higher pixel frequency than the 10D, but the 10D renders
    colors more accurately in an absolute sense.

    --
     
    JPS, Dec 2, 2003
  7. K2

    Mxsmanic Guest

    No, it removes 1/2 the green light, 3/4 of the blue light, and 3/4 of
    the red light. In other words, 2/3 of all light. That's a sensitivity
    loss of just over 1.5 stops.
    But it detects no red or blue at all. So only 1/3 of the light hitting
    the green pixel is detected. The same is true for the red and blue
    pixels.
    That's actually a better way of doing it, for maximum color fidelity.
    The Bayer pattern is a compromise that suits low-resolution sensors.
     
    Mxsmanic, Dec 2, 2003
  8. That is correct, but with only 25% of the advertised interpolated
    resolution. Non-interpolated resolution is all that matters.
    Compared to a non-interpolated 10.3MP sensor.
     
    George Preddy, Dec 2, 2003
  9. Simple, you are just learning about Foveon and informed people knew about it
    long ago. All the Bayer users in this forum who are still arguing about
    this actually thought the Foveon had only 3.43M sensors when these threads
    started, like an interpolaed Bayer sensor. That is how this whole thing
    started. Now, you fully understand it is a more efficient 10.3MP design,
    but you can't admit that 10.3MP is better than 6MP for fear of losing face.
     
    George Preddy, Dec 2, 2003
  10. It shows nothing, which speaks volumes.
    Then you continue to communicate your dissatisfaction with the 10D's image
    qualiy non-verbally- it is far more effective than anthing you could ever
    say.
     
    George Preddy, Dec 2, 2003
  11. If you try to convince yourself that 6.0/10.3 > 1.
     
    George Preddy, Dec 2, 2003
  12. So how come only Stigma bought Foveon's hype? A third tier maker doen't add
    any credibility.
     
    Darrell Larose, Dec 2, 2003
  13. Not on a per-pixel basis: every green sensor sees exactly the same intensity
    in a Bayer sensor as a green sensor in a 3-color sensor would see.

    There's no loss of sensitivity whatsoever.
    So? There's still no loss of sensitivity. Just chrominance resolution. Since
    you can't see chrominance resolution, you haven't lost anything you care
    about.

    More accurately, you haven't lost anything that you'd be willing to pay any
    cost in any other parameter to get back: B&W landscapes with a red filter
    are problematic with a Bayer camera, of course, but I no one would buy a
    monochome dSLR, as Kodak has already found out.
    It's a compromise that suits the human eye.

    David J. Littleboy
    Tokyo, Japan
     
    David J. Littleboy, Dec 2, 2003
  14. K2

    Rafe B. Guest


    All technology is a compromise. Best bang for
    the buck -- that's what engineering is all about.

    Bayer happens to match the physiology of the
    human visual system. In a nutshell: receptors
    for color are not capable of resolving much detail.

    There's a good reason why books are still
    printed with black ink on white pages, or why
    web sites with colored text on colored back-
    grounds are so annoying.

    In that regard, Foveon technology attempts to
    solve a problem that didn't really need solving
    - at least not as far as human vision goes.

    FWIW, JPEG encoding utilizes the same exact
    notion; one of the first steps is conversion to
    a Yuv/Lab color space, and decimating the color
    data (while preserving luminance.)



    rafe b.
    http://www.terrapinphoto.com
     
    Rafe B., Dec 2, 2003
  15. I don't want to get drawn into a flame war, but we should remember that so
    far we've seen only one version of the Foveon versus many versions of
    sensors with interpolated color. There are going to be more Foveons; it
    will be interesting to see how they evolve.
     
    Michael A. Covington, Dec 2, 2003
  16. 6/10 is less than one, of course. But spatial resolution is determined
    by the number of sample points, and the Sigma/Foveon camera has just 3.4
    million of them. It doesn't matter how many different measurements are
    made at each point; that doesn't improve spatial resolution any.
    And 6 is definitely larger than 3.4.

    However, it's not necessary to do any math. Simply looking at
    resolution tests shot by one of the 6 MP Bayer cameras shows that it
    resolves more real detail than the SD9.

    Dave
     
    Dave Martindale, Dec 2, 2003
  17. K2

    Guest Guest


    I am Preddy. I am Borg! You WILL be assimilated!


    No George, what matter is the final picture!!!
     
    Guest, Dec 2, 2003
  18. Correct, or a monochrome/unfiltered sensel exposed to green.
    There is only a very minor loss due to the close to, but not exactly, 100%
    transmission characteristics of the filter for the spectral band (there are
    no exact bandpass filter dyes). However that is something different than
    assuming a 2/3rd loss. The missing luminance and color is reconstructed with
    demosaicing, and it's rather accurate most of the time.

    Bart
     
    Bart van der Wolf, Dec 2, 2003
  19. face.

    No, it hasn't got 10.3M pixels, it has 10.3M sensors, stacked in three
    layers. Its spatial sampling resolution is only 3.43M pixels.

    Bart
     
    Bart van der Wolf, Dec 2, 2003
  20. K2

    Mxsmanic Guest

    But since no other information is gathered for that pixel, 2/3 of
    overall information is lost. And that is true for every pixel.
    Two thirds of sensitivity is lost, because each pixel receives only one
    color, instead of three.
    How can you receive only 1/3 of incident light with no loss of
    sensitivity?
    And luminance resolution, since 2/3 of it is gone. And total incident
    light, since 2/3 of it is filtered out.
    Only under specific conditions.
     
    Mxsmanic, Dec 2, 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.