Digital Photo Resolution & The Human Eye

Discussion in 'Digital Photography' started by javawizard, Jul 14, 2007.

  1. javawizard

    javawizard Guest

    Typical digital cameras have a resolution of about 8 megapixels these
    days. Just a few years ago, one, or three-megapixel resolution was
    typical. The human eye has a resolution of approximately 137
    megapixels. It's not quite a linear comparison, since our eyes have
    much higher resolution in the central area than at the edges. Still,
    you can imagine that soon cameras will be more sensitive than the
    human eye. - from the Technology section of
    javawizard, Jul 14, 2007
    1. Advertisements

  2. javawizard

    Scott W Guest

    137 sounds about right if you allow a person to look in any direction,
    resolution of the eye about 0.3 mr over 4*pi sr = 139MP.

    FWIW here is a photo that has about 172MP. (31 MB)
    Note the photo is too large to view in most browsers and so needs to be
    downloaded and viewed in a program Such are photoshop.

    The low res version can be seen here

    Scott W, Jul 15, 2007
    1. Advertisements

  3. 1. You appear to be spamming. So I have not visited that (your?)
    site, and will only reply to your posted 'info'. I'll be most
    interested to see if you return..

    2. There is NO way to sensibly compare such things. Your eye has high
    resolution over only a very small angle, as Scott pointed out. It's
    your brain that does the work of assembling the data collected and it
    can easily fool you into believing an entire scene is sharp even when
    it isn't. To prove that, FREEZE your eye on a word. While not moving
    your eye, can you read a line of text that is say, 6 lines down from
    the one you are staring at? If your eye has 137 Mp resolution, why
    can't you? And yet this unreadable text looks somehow 'sharp' - your
    brain does some clever tricks...
    mark.thomas.7, Jul 15, 2007
  4. It is very hard to compare resolution of the eye with a camera, either
    digital or film. The high resolution of the human eye is only in the
    foveal region (about 1 minute of arc there). But the fovea only
    encompasses about two degrees of vision. From that point the
    resolution degrades rapidly towards the periphery. If you take the
    total field of human vision and divide by an arc minute you very much
    overestimate human vision. The number of megapixels is far fewer.

    However, the eye scans a photo, looking at regions of interest with
    the fovea. The actual field of view the eye can scan is hard to fix-
    do we allow the head to move, or just the eyes?
    Don Stauffer in Minnesota, Jul 15, 2007
  5. The web site says no more that what is stated
    above, like everything else on the web site. It doesn't say
    how the derivation was made, nothing, not even a reference.

    For a little more detail, and derivations ranging from
    324 to 576 megapixels and greater, see:

    Notes on the Resolution and Other Details of the Human Eye

    I would say your 137 is quite low.

    Roger N. Clark (change username to rnclark), Jul 15, 2007
  6. javawizard

    John Turco Guest


    Soon? Not at that rate!

    John Turco <>
    John Turco, Jul 18, 2007
  7. javawizard

    Gladiator Guest

    Maybe your eyes but mine are 1mp at best.
    Gladiator, Jul 18, 2007
  8. Desert Dweller, Jul 18, 2007
  9. javawizard

    Ray Paseur Guest

    This topic brings to mind that quote about beauty: In the eye of the
    beholder (or beer holder, if you prefer). You can achieve acceptable
    fine art photographic enlargements with conventional 6MP cameras, but
    you can't read license plates from three miles away -- the optics aren't
    good enough. Specialized cameras can take on that surveillance work,
    but they rarely produce something you would hang on your walls.

    In my humble experience, it's the lens, more than the megapixels that
    matters. Also, the bit depth of the image seems to matter more than the
    megapixels when I make enlargements or adjustments to the original

    The exception to this broad generalization comes up when you haven't got
    the right lens or your original composition is not what you wanted, and
    you need to crop the image, then enlarge it for printing. In this
    situation more megapixels are better. But as a practical matter it is
    not easy to get enough megapixels, even by changing cameras, to make
    good enlargements from severe crops.

    Cameras record a two-dimensional view of the world, and image
    information increases much more slowly than the megapixel count. To
    double the dimensions of an image from a 6MP camera, you need to record
    24MP. That gets expensive!
    Ray Paseur, Jul 18, 2007
  10. javawizard

    Ron Hunter Guest

    Assuming that all humans have the same visual acuity is rather like
    assuming they all have the same blood pressure, or blood glucose level,
    or skin color.
    Ron Hunter, Jul 18, 2007
  11. It depends on whether you mean with scanning or without scanning the
    eye. The resolution of the fovea is high, but only covers 2 degrees
    field of view. We scan our eye to cover areas of interest. Now,
    because of that, if we talk about the whole scene that captures our
    interest, the number of pixels varies with the field of view of our
    "area of interest". Thus, I believe it is not really proper to
    compare number of pixels of eye and either film or a digital focal

    I believe it is valid only to account for visual acuity of both. That
    of course is dependent on the focal length of the lens on a camera, so
    it is still hard to compare the eye and a digicam. They just work too

    In camera system terms one is a scanning system and one is a staring
    Don Stauffer in Minnesota, Jul 19, 2007
  12. javawizard

    Gladiator Guest

    Yea, and that is what I meant when I said my eyes are only 1mp. My
    long distance viewing is not too bad but my macro capability is very
    Gladiator, Jul 20, 2007
  13. There is a simple way to look at the problem (pun intended):

    If you were standing and looking at a scene, then replaced
    that view with a print in front of you, what resolution would
    be required for you to be able to view the detail equal to
    the real scene? That includes naturally scanning your
    vision just like with the real scene.

    The quote of my page above is incomplete, as I derived different
    values considering different fields of view.

    This would be an interesting exercise for people to try:
    take an image of a location near your house, then quickly
    make a print and take it to the scene, and hold it up so
    the things in the print appear the same size as in the real
    scene. Now, can you see more or less in the print
    versus the real scene? How many megapixels is your print?
    How much of your eye's total field of view does that
    print appear (it will likely be a small fraction).

    Roger N. Clark (change username to rnclark), Jul 21, 2007
  14. [A complimentary Cc of this posting was sent to
    Desert Dweller
    My estimate was about 8MP. This assumes eye observing

    a) 3:2 format object;

    b) object is in horizontal format, so the visual angle is about
    50degree for the longer side;

    c) Lightning conditions which are not very bright (picture gallery).

    (Without being as detailed as I was, an estimate is not making any sense.)

    About "b": this is my non-scientific *impression* about how people
    view pictures/photos in the galleries: most of them would get "the
    general impression" from approximately such a distance.

    Note that one can get quite close to 8M optically-distinguished pixels
    with some 10MP sensors and good lenses.

    So any photo which is larger than 8M optically-distinguished pixels
    allows for "dive factor": if you have 24 such MP, the dive factor is
    2x: a person can "dive into the picture": come 2x closer, and STILL see
    no blur in details. (Or do likewise with looking at brighter light;
    this may be close to about 2x dive factor.)

    The dive factor 3 is very impressive; already 2 may stun some people
    who see it for the first time ;-).

    Hope this helps,
    Ilya Zakharevich, Jul 22, 2007
  15. javawizard

    BioColor Guest

    Another way to look at it (to unconfuse number of pixels and angular
    resolution) is to count the eye's pixels like camera manufacturers do.
    According to

    there are 6-7 million cones in the eye, each sensitive to one of three
    colors of light. Thus the eye is like a 2 megapixel color camera. We
    have 120 million rods that aren't sensitive to color. So the eye is
    like a 120 megapixel monochrome camera.

    The pixels are not evenly spaced, and thus the varying values of
    angular resolution.

    Considering eye movement in the calculation is a little like saying my
    D70 is a 100 megapixel camera because I could conceivably make a
    mosaic that big.

    BioColor, Jul 23, 2007
  16. David J Taylor, Jul 23, 2007
  17. javawizard

    BioColor Guest

    Thanks for those, David.

    Back in the last century (mid-80s) my assistant, David Oliver, and I
    made what was perhaps one of the largest color digital mosaics at that
    time for the National Geographic Atlas of North America (the first
    NatGeo atlas to include digital imagery). It is a mosaic of (if I
    remember right) 11 Landsat satellite scenes covering most of New
    England from New York City to central Maine, and from the Adirondack
    Mts to Eastport, Maine. I think it was about 14000 pixels on a side.
    with 80 meter resolution (about 1 acre).

    This was in the days of mainframes. The "giant" removable disk drives
    I was using held 175 MBytes each (on 6 platters). I had to process one
    of the three color bands at a time, using one drive for the input and
    one as the output. The biggest software challenge was that the
    different N-S "stripes" were done at different seasons. They had very
    different coloring. I developed an algorithm to do the color matching
    to hide the boundaries between scenes. I used a statistical measure
    and a "triangular warp" function that worked remarkably well. The only
    obvious boundary I ever found was a road that dead-ended because it
    was constructed after one of the stripes was acquired.

    The whole process took about 2 months with lots of through-the-night
    CPU cycles on the Cyber mainframe.

    I was fortunate to have a $250,000 Control Data/Ramtek device that
    would display a 1280x1024 image, so I could see what I was doing.

    There were no film recorders available that could write the full res
    image. I used an $250,000 Optronics film writer to write a 10x10
    ektachrome positive of a 2x2 reduction (one band at a time). NatGeo
    used that to print the atlas.

    Later the Smithsonian Air and Space Museum set up a remote sensing
    exhibit. They borrowed the film and made a 14-foot wall mural. As far
    as I know it is still hanging there.

    It would appear that there has been some progress on the digital
    mosaic front.

    BioColor, Jul 23, 2007
  18. javawizard

    The Real Bev Guest

    We visited the NG building in DC in the late 60s or early 70s and marveled
    at the photomosaic of the entire US on one of the walls. Do you know if
    it's still there or if it's been replaced with something better?

    Those were the days...

    Nobody needs to speak on behalf of idiots, they manage
    to speak entirely too much for themselves already.
    The Real Bev, Jul 24, 2007
  19. Fascinating, Duncan. You have a lot more patience than me!

    The individual images we get from European weather satellites today can be
    up to 11,000 pixels square - albeit with a ground resolution of 1km
    covering the whole disk (almost) from geostationary orbit. These images
    are sent every 15 minutes but, because of the bandwidth limitations from
    the satellite down to the ground, only half the scanned image width is
    sent (i.e. a strip 11136 pixels tall and 5568 pixels wide). Processing
    software for the full images takes about 124MB per plane - 372MB per RGB
    image. And on today's PCs two or three images that size can fit in memory
    at once...... Oh, and the users expect results at the press of a button.

    David J Taylor, Jul 24, 2007
  20. javawizard

    BioColor Guest

    On Tue, 24 Jul 2007 06:50:00 GMT, "David J Taylor"


    Thanks, David, for the numbers. At least they don't have to worry
    about spatial mosaicking. I did read a while ago that the registration
    through time for the "movie loops" presented lots of difficulty when
    it was first attempted. I think the driving force was the desire to
    put in state boundaries and roads for the weather reports on the
    nightly news.

    Right around the time I went to the opening reception for the
    Smithsonian exhibit (string quartet and lobster tails) the NatGeo
    Society was interested in doing something spectacular to commemorate
    their upcoming bicentennial. Having some experience now, we suggested
    doing over their lobby globe with Landsat satellite imagery instead of
    drawn maps. They didn't go for it. About a decade later the DeLorme
    mapping company just down the road from me did exactly that. It is a
    spectacular sight.

    BioColor, Jul 24, 2007
    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.