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

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

  1. Foveon does take 10.3M samples. 70% more than a 6MP Bayer.
     
    George Preddy, Dec 6, 2003
    1. Advertisements

  2. No. The layers are invisible to one another, so practically speaking, each
    one is the top layer. There is no depth involved from the photon's point
    of view--each hits a sensor. Bang. Ouch. I hit the surface. The R
    channel takes 3.43M spatially discrete samples, and so does G and B. No
    different that a 10.3MP Bayer, if such a wonderful thing existed.

    This isn't a hard concept, I'm surprised you are having so much trouble with
    it.
     
    George Preddy, Dec 6, 2003
    1. Advertisements

  3. K2

    JPS Guest

    In message <bqt99n$fff$>,
    (2268*1512*3)/(3072*2048) = 1.635, or 63.5% more sensors. However, they
    are placed three to a location, so the number of spatial witnesses is
    only 3.43M, or 54.5% of the 6.3mp Bayer.
    --
     
    JPS, Dec 6, 2003
  4. That sounds impressive, but it doesn't describe the Foveon.

    Bart
     
    Bart van der Wolf, Dec 6, 2003
  5. The problem we have with the concept is that it is wrong.


    Roland
     
    Roland Karlsson, Dec 7, 2003
  6. K2

    Azzz1588 Guest


    I just stopped at the yellow colored people.
    And the castle with the strange colors.

    And then george said the sd 9 had perfect color, the user set it
    himself.

    Does george see everyone in the world with yellow skin ????















    "Only a Gentleman can insult me, and a true Gentleman never will..."
     
    Azzz1588, Dec 7, 2003
  7. All sharp images from a 10D are... oh wait, no such thing, sorry.
     
    George Preddy, Dec 9, 2003
  8. Right, Foveon uses RRR instead of RGB.
     
    George Preddy, Dec 9, 2003
  9. All sharp areas of a Stigma SD-9/SD-10 is full of Lateral Chromatic
    Aberrations, which is a sign of poorly made lenses. Unfortunately nobody
    makes real lenses for the Toy Sigmas...

    Think of it is a Digital Holga...
     
    Darrell Larose, Dec 9, 2003
  10. No, it uses XYZ, where X, Y and Z are some imaginary
    names on the three layers color response. Then RGB is
    computed thus:

    R = T11*X + T12*Y + T13*Z.
    G = T21*X + T22*Y + T23*Z.
    B = T31*X + T32*Y + T33*Z.


    /Roland
     
    Roland Karlsson, Dec 9, 2003
  11. K2

    JPS Guest

    In message <br55gq$ob2$>,
    It takes 3 pixels for full modulation on a properly sampled image. The
    10D (JPEGs) and the conversion programs do not sharpen images as much as
    SPP does by default, so you're never going to see striking
    pixel-to-pixel contrast on a 100% view.

    That's the way it should be, and it's the only way to sample what is
    really there, in a wholistic sense.

    The SD9 images do not deliver everything that can be consistently
    sampled up to a certain frequency;, they contain little isolated dots
    and their contrast creates a false illusion of image sharpness. A Canon
    1Ds image downsampled to 3.43mp with nearest neighbor looks as sharp as
    an SD9 image, and looks sharper than the full 1Ds image at 100%, but it
    is a horrible image.
    --
     
    JPS, Dec 9, 2003
  12. K2

    Mark M Guest

    You may be the world's crummiest photographer...
    ....But you sure are one heck of a Geek.
     
    Mark M, Dec 10, 2003
  13. Ouch, that wasn't very nice of you. Especially since he's already on the
    record as saying he will never post an image to support his point of view,
    not under any circumstance.
     
    George Preddy, Dec 10, 2003
  14. Right, 10.3MP vs 6MP has nothing to do with it.
    No such thing. Sharpness is an emprical measure, an image is either sharp,
    or it is not sharp. SD9 images are sharp. 10D images are not sharp.
    So finally, after all this, you admit 10D images without downsizing are
    blurry and thus require downsizing of their 6MP-interpolated output files in
    order to become competive. Good job, that is a big step for you. We are
    moving in the right direction.
     
    George Preddy, Dec 10, 2003
  15. K2

    JPS Guest

    In message <br6o4d$cdl$>,
    It doesn't, and it's not true, regardless. The SD9 is a 3.43MP camera.

    The reason it appears sharper, pixel-for pixel, is that it doesn't scan
    the entire focal plane within the rectangle of the sensor. It scans in
    little holes, with most of the light absorbed or bounced by the vast
    area between photosites, and this guarantees lots of contrast at a
    1-pixel level if a sharp lens is used, even though the image itself is
    non-contiguous and compromised. The SPP software, by default, sharpens
    the images even further. If you kept comparing the SD9 output to
    greater and greater resolutions of bayer sensors at 100% zoom, the
    bayers would get softer and softer as you increased the MPs. The
    overall image would contain more and more detail, though, and would look
    sharper and sharper printed at the same size for the entire image.
    That's the forest that your viewpoint misses for the trees.
    The rasters output by the SPP software are very sharp, if the lens is
    sharp. A raster is not an image, though. An image is what is contained
    in the raster. The image is the message; the raster is the medium.

    A sound sample taken with a 6-bit 2000 Hz ADC in 100us windows (20% of
    the period), and no filtering on input or output will "sound" very
    bright. The "sound", however, is not an accurate recording of the
    original, and your SD9 is a graphic analog of such a sampling disaster.
    No; I say that a 10D file is softer out of the camera, pixel-for-pixel,
    than an SD9 file taken with the same lens. The whole 10D image,
    however, has better color accuracy at a lower frequency, and is sharper
    in real image texture, printed on paper. Both need to be reduced to see
    the entire image on the monitor, and the SD9 keeps all its aliasing
    artifacts as it shrinks the image.

    You are using the low resolution of computer monitors to the SD9's
    advantage, and this is only possible because you are too blind to tell
    the difference between artifact and recording detail.
    --
     
    JPS, Dec 10, 2003
  16. K2

    Guest Guest

    It takes 3 pixels for full modulation on a properly sampled image. The
    You are catching on Goerge. Its the lack of an AA filter and some SPP
    sharpening. Even if your camera did have 10.3MP which it doesn't.
    10D images out of the camera generally improve with some sharpening.
    SD9 images straight out of SPP look sharp, but you have been told time and
    time and time and time again why that is.
    He didn't say that, he said that if you downsample a 1Ds image to 3.43 MP if
    will look sharp like your alias artefact ridden, badly colored SD9 image.
    You can't even get the right camera - quite possibly you can't really read
    properly and ignore all the big words.
     
    Guest, Dec 10, 2003
  17. K2

    Mark M Guest

    Oops! My bad. -I thought I was talking to YOU, George...not Roland.
     
    Mark M, Dec 10, 2003
  18. K2

    pehache Guest

    ...And all complex 10D images have chromatic artefacts (this is true
    also for most of the other bayer-based cameras) .

    So what ?

    You can choose what you prefer: aliasing or chromatic artefacts :)
     
    pehache, Dec 10, 2003
  19. K2

    JPS Guest

    In message <>,
    Nonsense. Lack of chromatic resolution does not equal chromatic
    artifact. True, visible artifacts are uncommon on 6MP bayer DSLRs.
    --
     
    JPS, Dec 10, 2003
  20. Muhahaha ... good. I got somewhat worried there at first :)

    No George - I have not said that I never will post any photos
    to prove my point. I have said that I don't own any DSLR, so
    I have no photos (taken by myself) to prove my point. But
    fortunately, the net is full of them. You have yourself George
    provided some anti Foveon pictures. Nice of you - I just
    hope that Foveon likes your efforts as much as I do.


    Roland
     
    Roland Karlsson, Dec 10, 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.