Fuji claims.

Discussion in 'Fuji' started by Bill, Nov 26, 2003.

  1. Bill

    Bill Guest

    As far as the Fuji S602, it is a 3MP camera with a 3MP sensor. It is sold as
    such by Fuji and labeled as such on the box! At least as far as this model
    is concerned, Fuji is not scamming the public. Yes the Super CCD has a
    unique design, but it is a 3MP sensor. Now they do include a 6MP mode IF the
    user chooses to use it....and yes that would be interpolated. However, this
    is no different than anyone using any other camera, transferring an image
    into the computer and using their photo editing software to resample to a
    larger size if they want to. But Fuji included this option in their
    firmware...again...the OPTION. It's no different than other manufacturers
    who choose to include in the camera's firmware OPTIONS to enhance color,
    sharpness,
    etc. Personally, I prefer to do any of those things on the PC where I can
    see what I'm doing. I don't like to modify an image in the camera...not even
    cropping. I can always do that later.

    But my point here is that Fuji, at least with the S602, is not trying to
    fool anyone. They do not call it a 6MP camera nor do they advertise it as
    such. The box is clearly labeled as a 3MP camera. They do mention the 6MP
    mode as a feature only.

    Bill
     
    Bill, Nov 26, 2003
    #1
    1. Advertisements

  2. Bill

    This Old Man Guest

    That's one excellent camera. As soon as they drop below $300 on ebay I'm
    getting another.
     
    This Old Man, Nov 26, 2003
    #2
    1. Advertisements

  3. Bill

    JPS Guest

    In message <[email protected]>,
    I wouldn't treat it that way. If the camera uses the superCCD with this
    pattern:

    0 0 0 0 0
    0 0 0 0
    0 0 0 0 0
    0 0 0 0

    Then you *MUST* get the 6mp files from the camera to see everything the
    camera captured. A 3mp file from the camera *IS* a downsampling, and
    throws away detail.
    --
     
    JPS, Nov 27, 2003
    #3
  4. Bill

    JPS Guest

    In message <bq3qv6$6p1$>,
    There is no "hexagonal" anything in the superCCD. The pixel centers
    fall like one color of a checkerboard; a square grid rotated 45 degrees.
    The pixels themselves are octagonal.
    --
     
    JPS, Nov 27, 2003
    #4
  5. Bill

    JPS Guest

    In message <bq3qv6$6p1$>,
    It's not debateable. No file with anything less than 6mp can contain
    everything cpatured by a 3mp superCCD.

    A 3mp output file has absolutely no correlation to the 3mp sensor.
    --
     
    JPS, Nov 27, 2003
    #5
  6. Bill

    Mark Herring Guest

    At the risk of reigniting some very old debates.......

    The basic resolution of the S602 (ie the number of samples acquired by
    the CCD) is 3.1 Mpixels. You will find this in Fuji literature and in
    all the reviews.

    The 6Mp output file is obtained thru interpolation---AKA upsampling.

    The most consistent thing that I have seen about the Fuji issues is
    theri claim that the diagonal sensor arrangement increases resoluion
    in the horiz and vertical where it matter most. Skeptics have argued
    that the whole thing was just a trick to get better response to test
    targets.

    It is certainly true that the sampling pattern in the CCD can affect
    the resolution in certain axes---that may be what you are referring
    to.

    All this aside, the fact remains that the camera starts with 3.1Mp
    from the sensor---the 6Mp file is created later on in the processing.

    **************************
    Mark Herring, Pasadena, Calif.
    Private e-mail: Just say no to "No".
     
    Mark Herring, Nov 27, 2003
    #6
  7. Bill

    Bill Guest

    Why *MUST* you? There are only 3.1 million photodiodes on the sensor,
    regardless of the way they are arranged, and despite their octagonal shape.
    I am very pleased with my S602, but it is what it is.
    Bill
     
    Bill, Nov 27, 2003
    #7
  8. shape.

    The arrangement makes the difference, causing a must if you don't want to
    throw away actual samples. Try extracting all samples from a grid where each
    line is staggered by half a pixel. For a rectangular output grid you must
    double the number of cells, but only half of them are filled with the
    original staggered samples. Might as well fill in the blanks as the
    demosaicing is done...

    Bart
     
    Bart van der Wolf, Nov 27, 2003
    #8
  9. SNIP
    That's a bit misleading. The output file contains exactly the unchanged 3
    megapixels it captured which doesn't deteriorate the resolution, and
    interpolates the ones missing in the grid which doesn't add resolution. The
    original resolution is unchanged.
    Actually it is clever and a slightly better choice, because human vision is
    more sensitive (mentally) to horizontal/vertical features. It's not a
    coincidence that evolution caused that to be, because gravity causes more
    objects to have a more "horizon"tal or vertical orientation than anything
    else. Also movement is more likely to be either horizontal or vertical.
    With a slightly more involved (Fourier) analysis, it also is obvious that
    average image data has stronger horizontal/vertical frequency content than
    e.g. diagonal.

    Bart
     
    Bart van der Wolf, Nov 27, 2003
    #9
  10. Bill

    Leonard Guest

    Not so. Just invent a file format that doesn't assume horizontal
    and vertical addressing axes.

    - Len
     
    Leonard, Nov 27, 2003
    #10
  11. Bill

    JPS Guest

    In message <>,
    It's the only logical resolution to output in, given the pixel
    arrangement. You can not represent what the camera captures in a 3.1MP
    file.

    --
     
    JPS, Nov 27, 2003
    #11
  12. Bill

    JPS Guest

    In message <[email protected]>,
    This isn't a number-of-pixels game, it is sensors collecting data from
    *specific* locations.

    You can not shift around the data collected into another shape without
    distorting the image spatially, or losing some of the detail to
    resampling.

    --
     
    JPS, Nov 27, 2003
    #12
  13. No it is not. The Fuji uses a 3 MP sensor rotated 45 degrees. This
    increases the horizontal and vertical resolution (there are now 1.4
    times as many rows and columns of pixels) but decreases diagonal
    resolution. On output, you have a choice: use interpolation to add the
    other 3 million pixels needed to produce a conventional square raster,
    and keep the extra horizontal/vertical resolution, or downsample
    everything to 3 megapixels, losing the extra resolution.
    It's not hexagonal, despite what you might have read. It's a square
    grid rotated 45 degrees. It improves resolution, not just interpolation
    - but only for horizontal and vertical features.
    The point is that if you have the camera save a 3 MP file, it has
    discarded some information, and you can't get it back later by enlarging
    to 6 MP. So doing so after the fact is pointless. If you have the
    camera write a 6 MP file in the first place, it retains the extra
    resolution. There's a difference.

    Dave
     
    Dave Martindale, Nov 27, 2003
    #13
  14. And now you have a file that can't be displayed or edited by any
    standard software. Even to display it with special software, you have
    to upsample the image 40% larger because the display you use is
    undoubtedly a square raster. Unless you're suggesting building special
    hardware too...

    Dave
     
    Dave Martindale, Nov 27, 2003
    #14
  15. Bill

    pjp Guest

    I seem to remember in basic psych I took years ago that the reason most
    north americans (etc.) detect horizontal/vertical/90degrees, parallel lines
    etc. is because of the environment they've been brought up in, e.g.
    basically box shaped buildings. Studies showed that aboriginal people who
    grew up in more ecological (for lack of a better word) structures
    demonstrate much poorer ability at accuratly detecting the above type of
    examples, however they were much better at detecting other things, e.g.
    animals hiding etc.
     
    pjp, Nov 27, 2003
    #15
  16. Bill

    pjp Guest

    Exactly, should be obvious to anyone who lays it all out on graph paper.
     
    pjp, Nov 27, 2003
    #16
  17. You've missed the point.

    The Fuji sensor captures more information in the vertical and horizontal
    directions than a 3MP array can represent (and _less_ information in the
    diagonal directions).

    Since the 3MP Fuji sensors are rotated 45 degrees, rotating back loses
    information unless you upsample. And you have to upsample from the raw data.

    Since the tests that people use only test the vertical and horizontal
    directions in detail, the Fuji 6MP files appear to hold more information
    than the 3MP files.

    It's not unreasonable to call it a sleazy trick, but the 6MP files are
    justified, sort of.

    David J. Littleboy
    Tokyo, Japan
     
    David J. Littleboy, Nov 28, 2003
    #17
  18. Bill

    JPS Guest

    In message <Yfpxb.19213$>,
    It still needs to be displayed at 6mp, at some point, to see all it
    contains.
    --
     
    JPS, Nov 28, 2003
    #18
  19. Well, you could display a 3MP file at 45 degrees...

    David J. Littleboy
    Tokyo, Japan
     
    David J. Littleboy, Nov 28, 2003
    #19
  20. Chiropractors love that idea ;-)

    Bart
     
    Bart van der Wolf, Nov 28, 2003
    #20
    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.