Another Fuji scam

Discussion in 'Fuji' started by Mark Herring, Oct 10, 2003.

  1. Not necessarily. Most urban scenes have far more horizontal and
    vertical features than they have diagonal ones, and our eyes are better
    at *seeing* horizontal and vertical features than diagonal ones. So it
    does make some sense to rotate the sensor to maximize perceived image
    quality with a given number of pixels.

    The problem is that there is no standard file format for storing these
    rotated-raster images, and converting to a standard row/column raster
    doubles the image size. This imposes a cost that, for me, is too high
    for the improvement gained. But the rotated sensor *is* a technically
    valid approach, and might be seen as worthwhile by some photographers.

    This is completely independent of the approach used to market the
    camera. The Fuji 3 MP cameras are actually good 3 MP cameras by all
    reports; they just aren't comparable to 6 MP cameras.

    Dave Martindale, Oct 17, 2003
    1. Advertisements

  2. Mark Herring

    Mark Herring Guest


    I don't believe this for a minute. What is it about our eyes that
    would prefer a particular orientation? I got into this debate with
    someone a while ago (maybe Dave?). At one point, I took a resolution
    target and slowly rotated it--looking carefully to see if I could see
    ANY change with orientation. Nothing--not the slightest difference.

    The world in general has no preference for horizontal and vertical at
    the higher spatial resolutions---where we are most concerned about the
    performance of our sensors. What IS true is that the human eye-brain
    system can process large amount of info and can--for example--"see" a
    linear feature in the presence of very low signal to noise ratio. Our
    eyes evolved to allow us to spot dinner while on the hunt, or to study
    leaves, bugs, stars, and other things requiring high resolution. None
    of these would lead to a special orientation of sensors (rods and
    cones). In fact, does anyone have link to a high-res picture of the
    human retina? I'm betting that the placement of rods and cones is
    totally random.

    Does ANYONE have some pix from controlled testing comparing a
    "super-CCD" Fuji to some other camera? I was sent to one site by
    "Bill" (post in this thread), but the pictures were too low a
    resolution to see a difference. (And the targets were horiz and
    vertical bars---just what David says that Fuji was optimizing for!!)

    digital photos, more and better computers,
    and never enough time to do the projects.
    Private e-mail: Just say no to No
    Mark Herring, Oct 18, 2003
    1. Advertisements

  3. Mark Herring

    Mark Herring Guest

    No Fuji did not fire me.......

    Why do I care? Simple: I design and build digital camera systems for
    a living, and I am very well versed in the technology and
    specifications. In the consumer market, we have seen the evolution of
    a de-facto standard for specifying resoltuion--to wit: Megapixels or
    "Megapixel resolution". It is--in fact--not a complete definition
    since it only tells you the limiting resolution set by sampling
    theory. The true resolution depends on a host of other factors.

    Resolution in Megapixels, DOES however, offer a useful bench mark for
    comparing cameras. When someone like Fuji starts playing with the
    numbers, people are misled. This is bad for the customer, and bad for
    the industry.

    I am not out to save the world, but if I have knowledge that will help
    someone avoid making an uninformed decision, I am going to share it.
    If you ignore my information or advice---trust me---my feelings will
    not be hurt.

    Once again: The F700 is NOT a 6.2 Megapixel camera, according to
    prevailing standards. I can back that statement up with all the facts
    and theory that you want. On the off chance that the group has heard
    enough, please feel free to continue this by private e-mail.

    digital photos, more and better computers,
    and never enough time to do the projects.
    Private e-mail: Just say no to No
    Mark Herring, Oct 18, 2003
  4. Our visual system consists of both eye and brain parts. Although it
    might start out having no angular preferences, are you sure that a
    lifetime of living in environments with mostly horizontal and vertical
    lines couldn't produce better discrimination for them than diagonal

    In any case, I didn't make that up. I've seen the effect described in
    several different publications, and I even provided a reference to some
    of them one of the previous times this was discussed. You can choose to
    ignore me, or go looking for the reference yourself.
    Well, they can't be *totally* random because each requires a certain
    amount of space. But they're not in any sort of well-defined grid
    either. Still, the cones (and rods) are only the very front end of the
    visual system, and there are many perceptual effects that simply cannot
    be due to the sensors alone. They must happen in later processing, and
    there have been many experiments trying to figure out what that
    processing is.
    The usual review sites like dpreview should have an array of standard
    test images plus sample pictorial images. You could do your own

    But here's a very simple test that doesn't require a second camera. Get
    a 6 Mp image from a 3 Mp Fuji camera from one of the web sites. Pick
    one that looks sharp and well-focused. Reduce it in size to 71% using
    the highest-quality resizing algorithm you can find. You've got a 3
    megapixel image at this point. Now enlarge it back to its original
    pixel demensions, an enlargement of about 141%, again using the
    highest-quality resizing method you can find. Compare with the

    If it's true that the 6 megapixel output is "a scam", then the result of
    resizing down and back up should contain every bit of the original image
    detail (but possibly with either more or less contrast, depending on the
    resizing algorithm). On the other hand, if there is any detail in the
    original image that is lost by the double resizing, then the Fuji camera
    did capture and store information that a regular 3 megapixel camera
    could not do.

    (Signal processing theory predicts that the Fuji will have *better*
    horizontal resolution, but *worse* diagonal resolution than a normal 3
    megapixel camera. Which one looks better will depend on picture

    Dave Martindale, Oct 19, 2003
  5. Anyone reading this group regularly will know how many real pixels are
    in the camera's sensor. Just like they should know that the Sigma
    camera based on the Foveon sensor is a 3 megapixel camera, and in no way
    comparable to a 9 megapixel camera.

    However, you write as if what Fuji has done is a complete fraud, and
    there is no point at all in using a rotated sensor grid. Theory does
    not back you up on this, as long as you grant that a lot of people take
    photos that have predominantly near-vertical and near-horizontal lines.
    You don't even have to believe that the eye's contrast sensitivity
    function depends on orientation.

    There *is* some scientific justification for what Fuji implemented. I
    don't think they're justified in marketing a 3 MP camera as a 6 MP one,
    but that's a somewhat different issue.

    Dave Martindale, Oct 19, 2003
  6. Mark Herring

    Ray Fischer Guest

    Makes sense. Two smaller sensors each with half the dynamic range of a
    larger sensor would be just about what I'd expect.
    Ray Fischer, Oct 19, 2003
  7. Mark Herring

    Guest Guest


    I know what you're saying, but I will go back to my analogy of comparing
    diesel and petrol engines.

    Standard rectilinear CCD is a diesel engine.
    New Fuji Super CCD is a petrol engine.

    Both get quoted by engine capacity, but this tells you nothing much about
    the relative performance of the enigines. So Fuji has decided to quote the
    new camera as 6MP as that is the number of photosites on the sensor, and you
    can output an image from the camera with 6MP. This claim is only misleading
    if you make additional assumptions about the quoted "resolution". In this
    case you are applying an assumption that megagpixels is directly
    proportional to resolution and that this proportion is identical for every
    sensor type, which is just not the case in reality.

    Just as you wouldn't buy a car based purely on the engine size, you
    shouldn't buy a camera purely based on MP. Do you think that a 2000cc diesel
    car is the same as a 2000cc petrol car? Is a 3000cc BMW the same as a
    formula one car?

    You only know which is better for YOUR needs if YOU try them out.
    Guest, Oct 20, 2003
  8. Mark Herring

    Mark Herring Guest

    George--you are my hero!!

    Not only do you back me up, but you provide more data than I was aware

    Interesting experience at Good Guys(Calif--USA) today: I casually
    asked if the f700 was really 6Mpixel--given the low price? A very
    articulate employee (maybe the manager) instantly responded: Oh
    no--it is interpolated. I then pointed out that the price tag simply
    said 6Mp---right next to a Kodak advertising 4Mp for the same price.
    He very quickly pointed out that Good Guys will not sell a camera
    without explaining things such as this. He also very quickly
    acknowledged that Fuji has a past with these issues.

    So--score one for this store.

    While I'm here, Dave Martindale has provided some interesting leads on
    how diagonal sensors may offer some improvement. I acknowledge the
    need to follow up on this.

    Meanwhile, what we really need is pictures from the F700 and from some
    other cameras---taken from an actual scene. Anyone know if this
    exists---or an easy way to get it done?

    digital photos, more and better computers,
    and never enough time to do the projects.
    Private e-mail: Just say no to No
    Mark Herring, Oct 21, 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.