Film still better with colours than digital?

Discussion in 'Digital Cameras' started by Rich, Oct 28, 2005.

  1. Rich

    Rich Guest

    Seems like digitals have two problem areas;
    Bright scenes can really alter colours from reality
    and reds seem to be difficult to render as well as other
    colours. Examples might be a bright blue sky that has been
    overexposed to a degree and whose colour has shifted to the
    cyan and red flowers when compared to flowers of other colours.
    There is something odd about the
    texture of the red flowers. That said, has anyone shot, then
    digitized a good quality 35mm negative and compared it to the output
    from one of the $1000 digitals recently?
    I saw a comparison in a recent mag of Velvia versus digital and
    it seemed to bear this out, but the pictures were too small to
    be sure.
    -Rich
     
    Rich, Oct 28, 2005
    #1
    1. Advertisements

  2. Bright scenes aren't a problem: just expose correctly. Contrasty scenes are
    hard, but with careful exposure and careful RAW conversion, DSLRs do almost
    as well in the highlights and much better in the shadows.

    Reds are hard. I pull out all the red books (since the covers have subtly
    different shades) from my bookshelves every time I get a new digital camera
    and see how well I can persuade it to do, and it's always depressing. But
    film doesn't do any better.

    Every _quantitative_ comparison I've ever seen shows digital to be far
    better than film in color accuracy.

    Both measure and represent an infinite spectrum with three colors, and thus
    can be expected to do very similar things.
    Velvia is a serious joke in terms of color accuracy: it's designed to be
    extreme. The reds in particular are off the wall. That's why we landscape
    types like it. Also, it's an extremely high contrast material with almost no
    ability to handle scenes with a wide range of brightnesses.

    So Velvia is a funny film to bring up, since it is particularly bad at the
    two areas you mentioned to be of concern.

    David J. Littleboy
    Tokyo, Japan
     
    David J. Littleboy, Oct 28, 2005
    #2
    1. Advertisements

  3. [A complimentary Cc of this posting was sent to
    David J. Littleboy
    Actually, some digitals do it with 4 colors. This may be beneficial
    for "tricky" color balance problems.

    E.g., 3 colors are enough to obtain a honest representation of

    "what the eye can see in this lightening conditions".

    All you need is that color curves of eye cones can be well
    approximated by linear combinations of color curves of the sensels.

    However, consider the following question:

    "this shot is made in tungsten light. Now what the eye would see if
    it were in sun light?"

    This requires approximating the cone sensitivity curves *multiplied*
    by the quotient of spectral densities of tungsten and sun lights.
    Obviously, with 3 types of sensels, if you can answer one question
    good enough, you won't be able to answer the other one.

    But the more types of sensels you have, the more "color conversion"
    questions you can answer well.

    Hope this helps,
    Ilya
     
    Ilya Zakharevich, Oct 28, 2005
    #3
  4. Rich

    Jem Raid Guest

    The first time I did a print from a half way decent digital camera, it was a
    Minolta 2 Mp thingy, I was reminded of an old transparency film called
    Ferraniacolour. I always thought that this film was far and away more
    natural than any of its contempories which seemed to me to be far too bright
    and optimistic.

    Jem
     
    Jem Raid, Oct 28, 2005
    #4
  5. Rich

    Bill Funk Guest

    LOL! That's great! And true! Definite Sig material.
    [/QUOTE]

    Hmmm...
    Maybe then, you can explain it to me.
    eawckyegcy seems to be sayign that analog film use prodiuces no waste;
    that's what "the waste stream is O(number of pictures taken)" means,
    right?
    How does that work? I mean, there's obviously waste, so how does that
    work?
     
    Bill Funk, Oct 28, 2005
    #5
  6. Have you ever looked at that MF digital backs do with a full 16bits?
    I don't think so.


    ******************************************************

    "I have been a witness, and these pictures are
    my testimony. The events I have recorded should
    not be forgotten and must not be repeated."

    -James Nachtwey-
    http://www.jamesnachtwey.com/
     
    John A. Stovall, Oct 28, 2005
    #6
  7. Rich

    kctan Guest

    When electronic camera was invented, many said suck because they only know
    how to use mechanical one. When auto focus camera was invented, many said
    suck again because they doubt its accuracy without even trying one. Now
    someone jump the gun and said digital suck due to personal sub-standard test
    and judgment...saying that technology today is just too primitive to really
    match film, has never seen a real unbelievable digital print output side by
    side comparing to an analog one.
     
    kctan, Oct 28, 2005
    #7
  8. Rich

    Rich Guest

    Major issue is ground and sky shots. Unless you are shooting 180
    degrees away from the sun and on a day with dry air, chances are the
    sky burns out if you expose correctly for the land. The blue colour
    is definitely off if you try to bring the sky down to a normal
    brightness in post-processing. Varies with camera too.
    Yes, I was going by what one of the articles said. But what you say
    about liking a striking film for landscapes is true enough. Some of
    the landscapes you see in magazines shot by pros are really off the
    charts when it comes to colour. The neon green grass, etc, but it's
    not confined to film shots.
    -Rich
     
    Rich, Oct 29, 2005
    #8
  9. Rich

    Rich Guest

    I'm not familiar with the dynamics of the 20D on reds itself. What I
    find is that the "look" not so much the colour of red is off in
    digital images. It's sometimes imparts an irridescence that has to be
    carefully dealt with in post-processing to eliminate.
    I think everyone has a "favorite" setting for cameras concerning
    saturation, contrast, etc. I leave mine alone but I will
    experiment a bit with it.
    -Rich
     
    Rich, Oct 29, 2005
    #9
  10. Rich

    Bill Funk Guest

    Bill Funk, Oct 29, 2005
    #10
  11. If you are using a decent dSLR, RSE (or Adobe Camera Raw in PS CS2) will
    give you at least one whole extra stop in the sky. It's pretty amazing.
    Neither Canon's software nor Breezebrowser do, though. With a fat-pixel
    dSLRs (Canon 1Dmk2 or 5D) it's at least 1.5 stops.

    Shoot at ISO 100, and bright harsh sun simply isn't a problem. I shoot
    medium format film for my pretentious/pretty stuff, and the idea that film
    has more latitude than digital looks to me to be simply false in real life.
    (This is largely because film has essentially zero latitude in the low
    direction. Shadow detail has extremely poor acutance and resolution, so
    isn't usable.)

    David J. Littleboy
    Tokyo, Japan
     
    David J. Littleboy, Oct 29, 2005
    #11
  12. Rich

    223rem Guest

    Same thing, they're linearly related :)
     
    223rem, Oct 29, 2005
    #12
  13. Rich

    kctan Guest

    It is because Nikon was too ambitious then and wanted its user be able to
    use its old manual lenses on the auto focus body. In order to do so, many
    compromises hinder the design of a good auto focus camera body. Canon
    thought otherwise and re-design the auto focus lenses; throwing away the
    excellent FD series thinking that who would want to use manual focus when
    they have auto focus function. Eventually, Canon overtakes Nikon market.
     
    kctan, Oct 29, 2005
    #13
  14. Rich

    223rem Guest

    the two Ns are differ by a constant factor. So they go to inf
    equally 'fast'
     
    223rem, Oct 29, 2005
    #14
  15. [A complimentary Cc of this posting was sent to
    Roger N. Clark (change username to rnclark)
    Yes, thanks for correction.
    Would you be so kind to explain what you mean by "a linear function"?
    [All processes you consider are linear in usual sense of this word.]

    [It is the second time you claim that integration is a non-linear
    process. At least the second time when answering my posts. :-( You
    were already corrected once...]
    Nope. As I said, you need to match

    Eye_response(w) * Desired_lightening(w) / Actual_lightening(w).

    Here w is the light wavelength.

    Hope this helps,
    Ilya
     
    Ilya Zakharevich, Oct 30, 2005
    #15
  16. [A complimentary Cc of this posting was sent to
    Chris Brown
    Irrelevant. If you have an n-channel receiver, and can represent the
    spectral sensitivity curves of cones as linear combinations of
    sensitivity curves of the channels, all *is* tip-top. Your channels
    will (obviously) get positive readings; but the sign of the linear
    combination may be negative. Depend on what color space (encoding)
    you choose for the result.

    Or do you think that any linear combination of positive number is
    positive? ;-)
    The 700nm being the peak of red response looks very fishy, which makes
    the rest of the discussion suspicious too. Probably they meant 600nm
    (do not have my Marr at hand now...)?

    Yours,
    Ilya
     
    Ilya Zakharevich, Oct 31, 2005
    #16
  17. [A complimentary Cc of this posting was sent to
    Chris Brown
    I used http://cvision.ucsd.edu/basicindex.htm; it is not garish.

    What I could not find is the data for MacAdams ellipses. E.g.,
    pictures of MacAdams ellipses in CIELab coordinates...

    If one could locate tabulated data about penetration of monochromatic
    light into doped silicon, one could also resolve better the question
    on how many layers one needs in Foveon-like sensor to get good
    approximation to eye sensitivity. (Let me recall that my hunch from
    eyeballing the graphs is that 5 should be enough.)

    Hope this helps,
    Ilya
     
    Ilya Zakharevich, Oct 31, 2005
    #17
  18. [A complimentary Cc of this posting was sent to
    Scott W
    Right. As another clarification, let me just repeat the obvious:
    colors which appear the same under one lighting condition would not
    appear the same under another.

    Thus although 3 channels are enough for honest reproduction of colors
    as seen by the eye *at the moment of taking the photo*, if one needs
    to predict how the scene would look like under different lighting, one
    needs more channels than 3.
    Right. Likewise if you use 3 monochromatic lasers as the source of
    white light. ;-)

    (May be very real really soon now: it looks like light diodes may
    progress into primary sources of household lighting. And *if* it
    happens that a particular type of diod has efficiency with a sharp
    peak at a particular frequency, some manufacturer may lock the
    light diode into producing this particular frequency only [using
    same mechanism as for lasers, or another one].)

    But given "more traditional" sources of light, those with wide slowly
    changing spectral curves, very few channels (such as 6 or 7) can
    approximate the effect of such sources with pretty good accuracy
    (especially taking into account low sensibility of eye to color
    variations).
    No doubt for this!

    Yours,
    Ilya
     
    Ilya Zakharevich, Oct 31, 2005
    #18
  19. Rich

    Larry Lynch Guest

    It'll be a while before I can use my "O" key without guilt!!!!
     
    Larry Lynch, Oct 31, 2005
    #19
  20. [A complimentary Cc of this posting was sent to
    Roger N. Clark (change username to rnclark)
    a) 3-channel system is a particular kind of many-channels system. My
    argument is equally applicable to 3-channel and 100-channel systems.

    b) "Real world" uses non only 3-channel systems. Some digicams are
    4-channels.
    Did you *actually* read the article? Use "Find" button in your browser.

    The 700 nm wavelength ... was chosen because it is at the peak of
    the eye's red response

    Hope this helps,
    Ilya
     
    Ilya Zakharevich, Nov 1, 2005
    #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.