CFA demosaicing examples

Discussion in 'Digital Photography' started by eawckyegcy@yahoo.com, Feb 6, 2004.

  1. Guest

    Most people probably just look at what comes out of their camera.
    (Some even worship it!) But it is instructive nonetheless to examine
    different interpolation approaches. I just stumbled onto this:

    http://www.dmmd.net/Research/index.htm

    Click on "demosaicing" underneath "Image Processing". There are a
    small number of simulations of poor to excellent interpolation. (The
    "fence" is a whopper; look at how good the interpolator handled that
    chirp!) While you can tell the difference between the high-quality
    algorithm and the "original", note how carefully you have to inspect
    the results. Ask yourself the question: if the "DMMD" version came
    out of a camera, would you be able to tell a CFA was in use? Or
    suppose someone presented a field of (say) 30 images, a mixture of
    "original" and interpolates of undisclosed originals and a limited
    examination time; how many do you think you could correctly identify?
    (An interesting experiment! Hmmm...)
     
    , Feb 6, 2004
    #1
    1. Advertising

  2. <> wrote in message
    news:...
    SNIP
    > (An interesting experiment! Hmmm...)


    It is. That's why I did an experiment (
    http://www.xs4all.nl/~bvdwolf/main/foto/bayer/bayer_cfa.htm ) with an alpha
    version of their reconstruction implementation on an extremely critical
    image. The target used is sharper than one can expect from a camera with a
    lens and anti-aliasing filter, so I also simulated that influence.

    Bart
     
    Bart van der Wolf, Feb 6, 2004
    #2
    1. Advertising

  3. "Bart van der Wolf" <> writes:

    >It is. That's why I did an experiment (
    >http://www.xs4all.nl/~bvdwolf/main/foto/bayer/bayer_cfa.htm ) with an alpha
    >version of their reconstruction implementation on an extremely critical
    >image. The target used is sharper than one can expect from a camera with a
    >lens and anti-aliasing filter, so I also simulated that influence.


    If you'd like to simulate a real antialiasing filter, instead of using
    just Gaussian blur, do this:

    Copy the original image and paste the copy on top of the original,
    shifted one pixel left or right. Add the two images giving each 50%
    weighting. Then take the result and repeat the process again, making
    the offset one pixel up or down this time. This accurately simulates at
    least one particular type of AA filter. The one-pixel offset guarantees
    that the response at 0.5 cycles/pixel is exactly zero.

    You may still want to do a small amount of Gaussian blur, to simulate
    lens blur, before the AA filter simulation.

    Dave
     
    Dave Martindale, Feb 8, 2004
    #3
  4. "Dave Martindale" <> wrote in message
    news:c03vvl$no7$...
    SNIP
    > If you'd like to simulate a real antialiasing filter, instead of using
    > just Gaussian blur, do this:


    Thanks for the suggestion.

    Bart
     
    Bart van der Wolf, Feb 8, 2004
    #4
    1. Advertising

Want to reply to this thread or ask your own question?

It takes just 2 minutes to sign up (and it's free!). Just click the sign up button to choose a username and then you can ask your own questions on the forum.
Similar Threads
  1. jpc
    Replies:
    11
    Views:
    933
    Bart van der Wolf
    Nov 13, 2004
  2. Giox

    Demosaicing of Mg,Cy,Ye,G

    Giox, Oct 10, 2006, in forum: Digital Photography
    Replies:
    2
    Views:
    792
  3. rawebadvert3

    COVER LETTERS RESUME EXAMPLES AND CV EXAMPLES AVAILABLE

    rawebadvert3, May 25, 2007, in forum: Computer Support
    Replies:
    0
    Views:
    922
    rawebadvert3
    May 25, 2007
  4. rawebadvert3
    Replies:
    0
    Views:
    1,070
    rawebadvert3
    Aug 21, 2007
  5. rawebadvert3
    Replies:
    0
    Views:
    1,322
    rawebadvert3
    Mar 23, 2008
Loading...

Share This Page