Digital v Film

Discussion in 'Digital Photography' started by Simon Marchini, Jul 13, 2003.

  1. Simon Marchini

    Lionel Guest

    Has it occurred to you that perhaps the reason for that is that the
    experts have checked out that idea & found that it wasn't worth it?
    Yes it is. Firstly, by the maximum bandwidth of the video signal (which,
    IIRC, is about 6MHz for PAL), & colour is limited again by the brick
    wall frequency limits for the chroma signals. The limits for VHS
    recording are even lower still. It's utterly laughable to compare that
    level of 'quality' (either luma or chroma) to that of even a cheap P&S
    digicam, much less a DSLR. The fact that you're trying to do so just
    shows how desperate you are to try to confuse the argument.
    I won't be holding my breath waiting. All it takes for a Bayer sensor to
    beat such a system is for it to have triple (or better) the number of
    photosites. The Bayer sensor will always have a large cost advantage for
    the same image quality.
    (I've actually got an interesting idea on how to build a matrix sensor,
    very similar to a Bayer sensor, that might almost totally eliminate the
    aliasing/moire weaknesses, too. However, I won't be discussing the
    details publicly until I've done some research on feasibility, & talked
    to a patent attorney.)
    Lionel, Jul 17, 2003
    1. Advertisements

  2. Simon Marchini

    JPS Guest

    In message <bf2drk$o4l$>,
    It depends on what the operation is. Some things will be 199% as fast.
    Some will only be 105% as fast. Average will be around 190%.

    With only a 20% increase, as in the original context, you get almost all
    of it almost all of the time, as all my benchmarking suggests. Worst
    case scenario is a memory benchmark that's much bigger than L2 cache;
    that may benefit almost nothing at all.
    JPS, Jul 17, 2003
    1. Advertisements

  3. Simon Marchini

    Mxsmanic Guest

    Yes, but it is more likely to be a short-term marketing choice.
    Analog video systems are like film: resolution figures depend on the
    standards you apply, as there is no clear cutoff point.
    Mxsmanic, Jul 17, 2003
  4. Simon Marchini

    Rafe B. Guest

    Pretty sure. Send it to my work address:

    rafe b.
    Rafe B., Jul 17, 2003
  5. But still photography *also* has a far greater need for resolution above
    0.3 megapixels, while video does not. So far, increasing resolution has
    been more important.

    Three-sensor cameras using beamsplitter prisms have been around the
    video camera world for decades, using both tube and CCD sensors. It's
    off-the-shelf technology. If it made sense to use it in still cameras,
    it would be common. But, other than a few oddball or specialized
    cameras, it is not used. Are you suggesting that there's some sort of
    vast conspiracy to avoid using this superior technology? Or could it
    bee that camera manufacturers have found that the cost simply isn't
    worth the results, compared to a Bayer sensor.
    It's not fixed to a single number, but there is certainly an upper limit
    imposed by the analog bandwidth of the luminance channel.
    The Foveon is a different animal entirely. It's a lot more practical
    than a prism beamsplitter design because there is nothing but air
    between lens and sensor, so you can use standard lenses. The colour
    registration is inherent in the sensor design. The Foveon sensor might
    well take over from Bayer sensors in many cameras in the future. But
    you're changing the subject - you started out talking about 3CCD
    cameras, and that doesn't include anything with a Foveon sensor.

    Dave Martindale, Jul 17, 2003
  6. Is that a 55 Mbyte scan, or a 55 megapixel scan? The latter would be
    165 Mbytes, and that makes more sense in the context of your article.

    Yes, it would be startling at first. However, I'd be very interested if
    you would try the following experiment: Take one of those 55 Mpixel
    scans and reduce it to 50% using a good downsampling filter - bicubic at
    the least, or even better Lanczos. This will give you about 14 Mpixels,
    only slightly more than the 1 Mp image from the 1Ds. How does the
    reduced film image compare to the digital one? Does it still have most
    of the same information content? Is there a significant loss compared
    to the 55 Mp original?

    If the 14 Mp film image still looks a bit soft compared to the digital
    image, try doing a little bit of sharpening and see if you can make it
    equivalent. (Digital images usually have some sharpening done in the
    camera, while a film scan probably has none).

    I suspect that the 14 Mp downsampled image *will* still have all the
    most important information in it, while the difference between 14 and 55
    Mp will be visible but far less important. You may need to scan at 55
    Mp to capture all the image detail that the film recorded, but you don't
    always need to keep the higher-frequency part of that detail.

    Dave Martindale, Jul 17, 2003
  7. Irrelevant. You don't view images on a per-pixel basis, and if you scan

    film at a sufficient resolution to see the grain, its color accuracy

    is still worse on a per-pixel basis because the dye clouds' color

    is all over the place.
    Again, irrelevant. As soon as you transmit or record that signal,

    it's encoded in an analog or digital format that subsamples

    chrominance. The 3 chips help because both the luminance and

    chrominance resolution of video is so low to begin with.
    Irrelevant again. Do you have any evidence that Pop Photo's tests

    showing film's superiority to 1Ds files in lp/mm are incorrect?
    Obviously so are oil paints. Let me rephrase that since my point

    was apparently unclear. Portraiture plays to digicams' strengths

    and hides their weaknesses. You can get bigger good-looking

    enlargements from a given digicam pixel count with portraits than

    detailed landscapes, for instance.

    Russell Williams

    not speaking for Adobe Systems
    Russell Williams, Jul 17, 2003
  8. Right. 4000 dpi x 1.6 x 4000 dpi x 2.2 = 55MB.
    I do this with Provia 100F scans, and the resulting images are very nice. I
    suspect that Rafe's images will still be pretty grody, though. Negative film
    is really really horrendously bad.

    FWIW, (lightly sharpen, downsample 70%, lightly sharpen, downsample 70%,
    lightly sharpen) looks a bit better than (lightly sharpen, downsample 100%,
    lightly sharpen). IMHO.
    Actually, 2000 dpi scans with the Nikon 8000 look very good and get most of
    what's there as opposed to the 4000 dpi scan's nearly all.

    The high-frequency information recorded by film is recorded at such low
    contrast that it's pretty useless for imaging. Radical sharpening could
    help, except for the minor problem that grain noise prevents you from doing
    much of any sharpening at all. Film is a deeply and fatally flawed medium.
    (And film flatness is a _major_ problem for wide angle lenses.)

    David J. Littleboy
    Tokyo, Japan
    David J. Littleboy, Jul 17, 2003
  9. I've thought for some time that the subject and style of photos
    changes all these tradeoffs.

    The 16x24 print I've got from a 6mp digital file is in fact of a
    person (it's actually jewelry on a model, so I wouldn't call it a
    portrait, but I think it benefits in most of the same ways that a
    portrait does).
    David Dyer-Bennet, Jul 17, 2003
  10. SNIP
    The film can have some acutance added during processing (due to local
    depletion and migration of chemicals at high contrast edges) but how that
    exactly works out on the MTF of the scanner is largely unpredictable, but it
    will probably have a similar effect as in-camera sharpening, only more
    localized (like a mild edge masked USM).

    Bart van der Wolf, Jul 17, 2003
  11. Its a great scanner, I compared it to the scans done at a local
    prolab that were done with an Imacon, like I've always stated
    its more about the "operator" my scans are far superior for
    1 10th the price of the Imacon, mostly because I know how to
    optimize the 2450 and how to color balance.


    PS I think Tony's real problem is he doesn't know
    what he's looking for or at.

    Check out My Homepage at

    Support bacteria - they're the only culture
    some people have." -Stephen Wright
    Gregory W. Blank, Jul 17, 2003
  12. I can shoot alot of of 120 film & process it for 8K, Ok 5K
    thats the difference between it and my Bronica.

    Check out My Homepage at

    Support bacteria - they're the only culture
    some people have." -Stephen Wright
    Gregory W. Blank, Jul 17, 2003
  13. True intially I've spent a good deal on lab equipment, but the prints
    cost about 45ยข to make 8x10's myself.

    Check out My Homepage at

    Support bacteria - they're the only culture
    some people have." -Stephen Wright
    Gregory W. Blank, Jul 17, 2003
  14. Creative language from creative people.

    Check out My Homepage at

    Support bacteria - they're the only culture
    some people have." -Stephen Wright
    Gregory W. Blank, Jul 17, 2003
  15. Edge sharpening has always been a function of more contrast, likewise
    reduced contrast tends to show more grain especially in shadow areas of the film
    more sensitive films exhibit lengthened contrast gradients. This is a common occurance in photosensitive materials
    whether the technology exists or can be created to exist,..virtually
    that is to completely eliminate "this" occurance which is essence a function of physical
    nature remains to be seen. At this begining of digital imaging it seems that that companies
    are more or less working towards mimicking film exposures, if the far superior technology exists
    and is not being released, shame on them, is then just a game to capture dollars in as many routes as
    possible,...this is what I suspect (translated here,...... what I know) .

    Check out My Homepage at

    Support bacteria - they're the only culture
    some people have." -Stephen Wright
    Gregory W. Blank, Jul 17, 2003
  16. Of course, I meant "the 11 Mp image from the 1Ds".

    Dave Martindale, Jul 17, 2003
  17. Simon Marchini

    Mxsmanic Guest

    First, scanning at a resolution high enough to actually see dye clouds
    is far easier said than done (I've never seen that myself). Second, the
    color is not "all over the place," it just contains a certain amount of
    random noise.

    If the final image is 1 megapixel, why would ten megapixels be better
    than five as a starting point?
    I have a low opinion of Popular Photography generally.
    Mxsmanic, Jul 17, 2003
  18. Simon Marchini

    JPS Guest

    In message <bf6iu1$dhd$>,
    I really only see that as being worthwhile if technology hits a brick
    wall on crowding pixels into a given space, but can still go three deep,
    or if lenses can't resolve the added luminance resolution of higher-
    resolution bayer sensors. With finite pixels-channels as a limiting
    factor, it seems that the pixels are best spent in a bayer-like pattern.

    If I take a sharpened, downsampled image (downsampled enough to lose all
    bayer-based interpolation), I can pixellate the chrominance channel up
    to 4*4, with only a hint of lost color definition. If I pixellate the
    luminance channel, even 2*2 severely destroys the image.
    JPS, Jul 17, 2003
  19. Simon Marchini

    JPS Guest

    In message <bf6nrj$hko$>,
    Talk about compression!
    JPS, Jul 17, 2003
  20. (Gregory W. Blank) wrote in :
    While the reviews are quite good for your scanner, Gregory, and I'm sure
    it's capable of fine results, if I *were* to use an Epson Perfection 2450
    as the source of my scans to compare with my digital images, you know very
    well that the anti-digital people here would laugh their asses off and
    proclaim the comparison to be invalid because of an inferior scanner. You
    KNOW they would do that.
    Tony Whitaker, Jul 17, 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.