Fuji's 5th generation Super CCD HR sensor

Discussion in 'Digital Photography' started by Clint Kirk, Jan 16, 2006.

  1. Clint Kirk

    Clint Kirk Guest


    Can anyone shed any light on how this new sensor works? Take, for
    example, the 9MP one found on their E900 and S9000 (or S9500) cameras.
    Fuji says this is 9 million effective pixels, and 9 million recorded
    pixels, therefore there is no interpolation (such as there was on their
    4th generation Super CCD HR sensors.)

    But how can there not be interpolation? The sensor is still tilted at
    45 degrees to the horizontal, so how can you produce 9 MP in a
    horizontal/vertical orientation without interpolating and/or

    Clint Kirk, Jan 16, 2006
    1. Advertisements

  2. Every bayer sensor does interpolation of colour information, where every
    pixel in the resulting image is interpolated out of the colour
    information in that and surrounding pixels on the sensor. Fuji's Super
    CCD does the same, the only difference being that the final pixel
    doesn't have a direct positional relationship with 1 individual pixel on
    the sensor. This is why previous SCCD's cameras had their standard
    resolution and also a double resolution. In the double resolution mode,
    each pixel was either interpolated or had a direct relationship with an
    individual pixel. The end result is that for many images, an increase in
    perceived resolution can be gained.
    Graham Fountain, Jan 16, 2006
    1. Advertisements

  3. Clint Kirk

    ASAAR Guest

    I can't shed much light on how Fuji's sensors work, but there's
    interpolation and there's interpolation. In fact there needs to be
    interpolation in *all* Bayer sensors, but that's of a different
    kind. What Fuji means is that with their earlier generation Super
    CCD sensors, you could take a particular one having, say, 6 million
    actual pixels and have the camera set to save image files containing
    only 6mp. Or you could enable interpolation which would then create
    12mp image files. Naturally, the file size would double, and it
    wouldn't come close to providing the detail that a true 12mp sensor
    would. According to reviews, it did help slightly, maybe being
    equivalent to a 6.3mp sensor, but the amount of improvement was
    dependent on the subject matter. I recall some reviewers saying
    that it helped least when dealing with horizontal and vertical
    linear objects and did better when they were at some intermediate
    angle. If the E900 and S9000/S9500 used the same (or similar) 9mp
    sensors and created interpolated output files, they would be twice
    as large, containing 18 million interpolated pixels, which few
    people would find worth dealing with, cutting in half the number of
    pictures you could store on flash cards, and slowing the operation
    of photo editing on the computer. Sometimes enormously so, if the
    computer's memory resources are borderline . . .
    ASAAR, Jan 16, 2006
  4. I am more interested in knowing what have they done in the newer 1/1.7"
    Super HR CCD sensor used in the Fuji F10/11 that makes it such a
    low-noise sensor compared to other P&S cameras in its competition.

    - Siddhartha
    Siddhartha Jain, Jan 16, 2006
  5. Clint Kirk

    Clint Kirk Guest

    Thanks to everyone for the answer's, especially to Graham Fountain who
    pointed out the interpolation done by all Bayer (i.e. mosaic filter)
    sensors. Actually, I was trying to simplify my question by assuming,
    for now, that the CCD was monochrome, and once that was answered I
    would interpolate (no pun intended) the answer to include the colour
    information by thinking about the effects of the mosaic filter.

    In the end, as Graham pointed out, all Bayer sensors interpolate anyway
    (they interpolate two out of the three colour values in each pixel).
    So, since every pixel is interpolated in all such sensors (not just
    Fuji's ones) then it's really not such a shame that every pixel in
    Fuji's new sensor is also interpolated. Is my understanding correct?

    Clint Kirk, Jan 16, 2006
  6. Clint Kirk

    Bill Funk Guest

    There's a difference between "pixel" and "color in a pixel."
    Bayer-filtered sensor cameras interpolate the chroma in pixels, not
    the pixels.
    In the Fuji, the actual pixels are interpolated, because the pixels in
    the sensor and in the final image are entirely different. There's no
    way the sensor's sensel layout can be made into a rectangular format
    without interpolating the pixels themselves.
    This isn't intended to be a comment or complaint about Fuji's system,
    just a comment that the pixels themselves are interpolated.
    Bill Funk, Jan 16, 2006
  7. In recent Fuji sensors, half the sensor locations have relatively large
    sensors that have good light sensitivity with high effective ISO to
    capture detail in shadows, and the other half are small sensors with
    low ISO to capture highlight detail without burning out. So a "9 MP"
    sensor is really 4.5 MP of high-sensitivity locations plus 4.5 MP of
    low-sensitivity locations.

    In earlier versions of this, the little low-ISO sensors were placed
    close to the high-ISO sensors and are probably better regarded as a pair
    located in one place. They were located in a rotated square grid, so
    (as usual with Fuji), needed interpolation to get a row/column raster
    for the image file without losing any resolution. So it was a 4.5 MP
    sensor interpolated up to 9 MP, as usual, with a nearly-coincident 4.5
    MP second sensor for highlight information.

    The latest designs seem to place the small sensors halfway between the
    large ones, like the white and black squares on a checkerboard. So, at
    the least, each output pixel is actually located at a place where there
    is a sensing element on the sensor. But half of them capture only
    high-brightness components and the other half capture only
    low-brightness components, so interpolation is still necessary to
    produce a full dynamic range at each pixel.

    On the other hand, there now are 9 million separate measuring locations
    in a "9 MP" sensor. In theory, the sensor could have resolution
    equalling that of a standard 9 MP sensor with 9 million identical
    measuring cells, at least for the midtone portions of the image that
    can be handled by both the small and large cells. So it would be very
    interesting looking at some resolution tests.

    Dave Martindale, Jan 17, 2006
  8. Clint Kirk

    Clint Kirk Guest

    That's very interesting. It got me thinking what it would look like
    once you've put the colour filter on top.
    Clint Kirk, Jan 18, 2006
    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.