anti-aliasing and scanners, what's the deal?

Discussion in 'Digital Photography' started by Paul Rubin, Sep 23, 2006.

  1. Paul Rubin

    Paul Rubin Guest

    For example, say I have an 8x10" view camera transparency with lots of
    fine detail. I scan it in a flatbed scanner which uses a linear CCD
    sensor at 300 dpi.

    Do these scanners have any AA filters? If not, don't I get aliasing?
     
    Paul Rubin, Sep 23, 2006
    #1
    1. Advertisements

  2. Dunno about 300 ppi, but there is a phenomenon known as "grain aliasing" in
    which the grain in negative films is exacerbated due to aliasing. This was a
    nasty problem with 2700 ppi scanners and consumer color negative films, and
    is less of a problem with 4000 ppi scanners. It's even less of a problem
    than it looks at 4000 ppi (high-res scans of negative films look pretty
    ugly*), since 9x is about the limit for quality enlargements (please, note
    the word "quality" in there) and you can be fairly aggressive with noise
    reduction at 4000 ppi and then downsample to 2700 ppi for printing at 300
    ppi.

    The folks at Minolta had the right idea of scanning at 5400 ppi, since
    that's a full factor of two overscan of my desired target final resolution.
    A pity it doesn't scan MF. Sigh.

    *: See the following page for examples.
    http://www.terrapinphoto.com/jmdavis/

    David J. Littleboy
    Tokyo, Japan
     
    David J. Littleboy, Sep 23, 2006
    #2
    1. Advertisements

  3. Aliasing is only a big problem if you have periodic or repetitive
    detail at high spatial frequencies. Then indeed you can see aliasing
    artifacts. The most common place this shows up is when scanning bar
    chart lens testing patterns. Fence lines and such can create this in
    real images, but it is generally not much of a problem.

    Some scanners do include anti-aliasing filters. Also the "halftone
    removal" or descreening functions may also do some anti-aliasing.

    Grain does not normally create much of an aliasing problem because it
    is aperiodic- that is it has no really strong single spatial frequency
    content.
     
    Don Stauffer in Minnesota, Sep 23, 2006
    #3
  4. The native scanner sampling density is much higher than that, in the
    3200-6400 ppi range. Don't confuse that with the output resolution,
    because that is reached by resampling the scanner data.
    No AA filter, therefore any detail smaller than half of the sampling
    density will be subject to aliasing, e.g. film grain aggregates /
    dye-clouds. The detail in film itself is of relatively low modulation
    due to the interaction between the camera lens and the film emulsion,
    and image detail is somewhat irregular due to graininess, so the image
    aliasing is not a real issue with most films.
     
    Bart van der Wolf, Sep 23, 2006
    #4
  5. Paul Rubin

    Paul Rubin Guest

    Are you serious? I'm talking about an ordinary cheap desktop
    page-sized flatbed scanner, not a film scanner or some high end Epson
    Perfection scanner. They are usually specified at 300 or 600 dpi and
    I thought they got higher density by software interpolation.
    OK, thanks, this (and similar posts from other folks) clears up the
    mystery.
    I wonder what happens when you scan pages with text on them (small
    print). These do have sharp edges and therefore high frequencies, but
    maybe not over large areas.
     
    Paul Rubin, Sep 24, 2006
    #5
  6. I'm talking about flatbed film scanners (with a light lid for
    transparent film), not a simple document scanner. Reasonable document
    scanners, e.g. the Canon Lide types, typically have 600-1200 ppi
    sampling densities and are good enough for documents and half tone
    images as found in process printed matter.
    Yes, anything higher than the native sampling density is interpolated.

    SNIP
    A lot depends on the paper they are printed on. Very high quality
    type-setters on glossy paper can achieve 1200 dpi or more, but there
    is some spreading of ink (dot gain). Yet any detail smaller than half
    the sampling density wil to some degree cause aliasing artifacts which
    are by definition larger (more visible) than the actual structures
    causing it.
     
    Bart van der Wolf, Sep 24, 2006
    #6
    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.