Color Depth and Repro from digicam vs film

Discussion in 'Digital Cameras' started by Chris P in PA, Dec 29, 2003.

  1. I have seen much discussion on megapixel, and what you need/want to
    use for a given purpose.

    I have read some of the bayer-foveon debate.

    What i have yet to see discussed - anywhere at all - is color
    reproduction in digital cameras versus film...

    I have read about scanners that do 24 and 48 bit, and seen discussions
    here about editing in 16 bit, etc.

    got me to thinnking...

    how many shades/colors does 35mm color film record?
    How about digicams?

    more than 256, but 32,000? 250,000? 1 million?
    how many can the human eye discern?
    Chris P in PA, Dec 29, 2003
    1. Advertisements

  2. Billions. If we consider just the dynamic range of color film, then
    generally 10 to 12 bits per color (R x G x B = 30 to 36-bits total) is
    the digital equivalent to film. (This is based on info from
    Sinar-Bron, the camera equipment manufacturer, and Minolta.)

    If the camera samples at 8-bits per color, then it has a palette of 16.7
    million colors to choose from. A sample of 10-bits per color is 64
    times that, and 12-bits per color is 64 times that. However, most
    consumer digitals reduce that to 24-bit (8-bits per color) for output,
    which is more than enough for web, e-mail and snapshots, but not for
    "serious" photography.
    The human eye can "see" billions of colors, but can only discern about
    300,000 at any one time, which is not a problem, since most
    photographs, digital or silver-based, (or real world scenes for that
    matter) only have 10s of thousands of colors chosen from a palette of
    millions or billions, depending on the bit depth of the sampling.
    Stefan Patric, Dec 30, 2003
    1. Advertisements

  3. So, at least in B&W, most any digicam can outdo tri-x-pan? If not in
    resolution (dpi) then in shades of gray that can be captured?
    (assuming proper exposure in both cases)

    I would tend to think that nature can produce an infinite variety of
    colors and lighting that adds to that number, so the actual colors (or
    shades of gray) are essentially unlimited. I know the human eye can
    see far less range, and can probably discern even fewer total colors
    (at some point blue becomes blue purple, but is that is in 30 steps or
    300,000, and what can we see?)

    I know that in nature motion is linear. in film 24 fps and video 30fps
    appears to the human eye as continual motion. Just curious as to
    whether there is something similar regarding seeing color (and some
    folks are color blind so there MUST be some limit).

    I am generally familia rwith how film records images, but have no idea
    on the color range, or color resolution. I always here for best
    results, use slide film though.

    just wondering where the digicams sat in all this. Scanners are up to
    48 bit color depth, more than film.

    Chris P in PA, Dec 30, 2003
  4. What is your definition of "can be measured"? I suspect you're using a
    different definition of dynamic range than photographers usually do.
    If you limit yourself to the central portion of a film's characteristic
    curve, where the slope is more or less constant, you will indeed have
    only 5 or 6 stops of exposure range. But the film characteristic curve
    has shoulder and toe regions where the slope is reduced but not yet
    zero. Exposure in these regions produces *some* change in density for
    a change in exposure, and a photographer would consider these zones
    part of the dynamic range of the film because some information is
    captured. Using this definition, the dynamic range of a negative film
    can be 10 stops or more (greater than 1000:1)
    Since you haven't told us how your measurements were made, how can
    anyone do better? My own opinions aren't based on my own lab
    measurements, but on published film data. For example, if you look at
    Kodak's datasheet for Tri-X at
    and go to page 8 where the characteristic curves are shown, you'll see
    that the log E (X axis) range over which the slope is non-zero is 3.0 or
    greater. This is a base-10 log, so a 3.0 log E range is an exposure
    change of 10^3 = 1000. And these graphs don't show the ultimate drop in
    contrast at the shoulder, either - the film has even more highlight
    capture range than this.

    Or look at the typical film curves in "Photographic Materials and
    Processes" by Stroebel, Compton, Current, and Zakia. Their graphs cover
    a log E range of about 3.2.

    So I think you're understating the dynamic range over which film
    captures some information.

    Another issue is that, even if a scene has only a 64:1 dynamic range,
    that doesn't mean you can get away with quantizing the intensity to only
    6 bits. What a photographer really wants is an encoding that is fine
    enough so that you can't see the discrete steps between adjacent codes,
    even at the darker end of the intensity range. This requires more bits.

    For example, suppose we're using a linear encoding of intensity into
    pixel value. Also suppose that the eye can't see a step change that's
    smaller than about 3% in the dark end of the dynamic range of the image,
    which has a total range of 64:1. To keep the largest step change to
    about 3%, adjacent pixel codes must not differ by more than that in
    intensity. This requires about 5 bits to represent the dark end of the
    dynamic range, so that adjacent values would be (for example) 31 and 32.
    But the brightest part of the image is 64 times as bright, so its code
    value must be 2048, which requires 11 bits total. Basically, of the 11
    bits, 6 bits is needed for the 64:1 dynamic range, and the other 5 bits
    are needed to keep the step size small. Actual images would use pixel
    values in the range 32-2047, not ones below 32. If you used only 6 bits
    for the same image, adjacent pixels would have to be assigned values of
    1 or 2, a 100% change in intensity.

    In practice, we normally use a so-called gamma corrected encoding, which
    makes the math more complex. And the eye's ability to see a step change
    ranges from something less than 1% at high brightness to many percent in
    dark areas of an image. But the analysis of how many bits you need is
    still much the same: you need to keep the step size between adjacent
    codes small enough at the lower end of the useful dynamic range, and
    that always requires more bits than simply encoding the dynamic range.

    Dave Martindale, Dec 30, 2003
  5. Chris P in PA

    DJ Guest

    I have a lot of experience in making quantitative light intensity
    Fascinating! Can you describe approximately how those test are done? I'd guess
    you'd photograph a grey scale step wedge (is that the right term?) then use a
    densitometer to read the negative? Or directly expose a test strip with varying
    amounts of light?

    What mechanisms limit the dynamic range of film?

    Is there any limit to the number of steps that can be reproduced within that
    min/max range? I can't imagine a film photo ever exhibiting quantization
    artifacts like you sometimes see on digital.
    DJ, Dec 31, 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.