Film is not linear ?

Discussion in 'Digital Photography' started by Alfred Molon, Oct 6, 2004.

  1. Alfred Molon

    Alfred Molon Guest

    I got this impression when processing the scans of my brother's slides.
    Very strange colours and brightness levels. While CCDs are linear
    devices, where the output (before gamma) is a linear function of the
    light level, I get the impression that film does not perform like that
    and that the response curve is far from linear. Is this the case?
    --

    Alfred Molon
    ------------------------------
    http://groups.yahoo.com/group/Olympus_405080/
    Olympus 5060 resource - http://myolympus.org/5060/
    Olympus 8080 resource - http://myolympus.org/8080/
     
    Alfred Molon, Oct 6, 2004
    #1
    1. Advertising

  2. Alfred Molon wrote:
    > I got this impression when processing the scans of my brother's slides.
    > Very strange colours and brightness levels. While CCDs are linear
    > devices, where the output (before gamma) is a linear function of the
    > light level, I get the impression that film does not perform like that
    > and that the response curve is far from linear. Is this the case?



    I'll take you word that CCD's are linear. Film I know is not linear.
    Just look at the response curves.
    --
    Joseph E. Meehan

    26 + 6 = 1 It's Irish Math
     
    Joseph Meehan, Oct 6, 2004
    #2
    1. Advertising

  3. Alfred Molon

    Jeremy Nixon Guest

    Alfred Molon <> wrote:

    > I got this impression when processing the scans of my brother's slides.
    > Very strange colours and brightness levels. While CCDs are linear
    > devices, where the output (before gamma) is a linear function of the
    > light level, I get the impression that film does not perform like that
    > and that the response curve is far from linear. Is this the case?


    Not linear at all. Kodak used to publish little pamphlets with all the
    information about each of their types of film; they probably still do.
    One of the things you get there is the response curve, which is far from
    linear, even taking gamma into account.

    --
    Jeremy |
     
    Jeremy Nixon, Oct 6, 2004
    #3
  4. "Alfred Molon" <> wrote in message
    news:...
    >I got this impression when processing the scans of my brother's
    >slides.
    > Very strange colours and brightness levels. While CCDs are linear
    > devices, where the output (before gamma) is a linear function of the
    > light level, I get the impression that film does not perform like
    > that
    > and that the response curve is far from linear. Is this the case?
    > --
    >

    Isn't that what the gamma curve's all about? The non-linear response
    of film!

    Jim.
     
    James Silverton, Oct 7, 2004
    #4
  5. Alfred Molon

    Jim Guest

    "Alfred Molon" <> wrote in message
    news:...
    > I got this impression when processing the scans of my brother's slides.
    > Very strange colours and brightness levels. While CCDs are linear
    > devices, where the output (before gamma) is a linear function of the
    > light level, I get the impression that film does not perform like that
    > and that the response curve is far from linear. Is this the case?
    > --

    No film is not linear. Reciprocity failure is just another description of
    non-linearity.
    Jim
     
    Jim, Oct 7, 2004
    #5
  6. Alfred Molon <> writes:
    >I got this impression when processing the scans of my brother's slides.
    >Very strange colours and brightness levels. While CCDs are linear
    >devices, where the output (before gamma) is a linear function of the
    >light level, I get the impression that film does not perform like that
    >and that the response curve is far from linear. Is this the case?


    Definitely not.

    To a first approximation, over the main portion of its response curve,
    film behaves approximately like

    transmitted_light = some_constant * (exposure)^(gamma)

    where "^" means exponentiation. This is a power function
    whose effect depends greatly on the value of gamma.

    For B&W negative film, gamma is about -0.6. The negative gamma says
    that the process is a negative one: brighter light gives a darker image
    on the film. The 0.6, being less than 1, means that a certain contrast
    range in the scene is compressed to a smaller range on the film. For
    example, a 5-stop range (32:1) in the scene becomes only a 3-stop range
    (8:1) on film.

    Colour negative film has a gamma of about 0.5 or 0.55. The three layers
    are slightly different in gamma, and the "constant" is very different
    for the 3 layers because of the orange mask.

    Transparency film has a gamma of about 1.5. The positive value means
    it's a positive process, and 1.5 says the slide has a higher contrast
    than the original scene.

    There is special internegative film which has a gamma of -1.0 or close
    to it. This neither increases nor decreases contrast, but it still
    gives a negative image. Copying onto internegative film *twice* gives a
    reasonable copy of the original.

    A linear film would have a gamma of 1.0, but I know of no such film.

    Also, all the above applies only to the middle of the response curve.
    There are usually "shoulder" and "toe" regions with lower contrast for
    the shadows and extreme highlights respectively. And the centre section
    really isn't exactly straight either.

    Dave
     
    Dave Martindale, Oct 7, 2004
    #6
  7. Alfred Molon

    Alfred Molon Guest

    In article <ck2c6g$kre$>, says...

    > Definitely not.
    >
    > To a first approximation, over the main portion of its response curve,
    > film behaves approximately like
    >
    > transmitted_light = some_constant * (exposure)^(gamma)
    >
    > where "^" means exponentiation. This is a power function
    > whose effect depends greatly on the value of gamma.
    >
    > For B&W negative film, gamma is about -0.6. The negative gamma says
    > that the process is a negative one: brighter light gives a darker image
    > on the film. The 0.6, being less than 1, means that a certain contrast
    > range in the scene is compressed to a smaller range on the film. For
    > example, a 5-stop range (32:1) in the scene becomes only a 3-stop range
    > (8:1) on film.
    >
    > Colour negative film has a gamma of about 0.5 or 0.55. The three layers
    > are slightly different in gamma, and the "constant" is very different
    > for the 3 layers because of the orange mask.
    >
    > Transparency film has a gamma of about 1.5. The positive value means
    > it's a positive process, and 1.5 says the slide has a higher contrast
    > than the original scene.
    >
    > There is special internegative film which has a gamma of -1.0 or close
    > to it. This neither increases nor decreases contrast, but it still
    > gives a negative image. Copying onto internegative film *twice* gives a
    > reasonable copy of the original.
    >
    > A linear film would have a gamma of 1.0, but I know of no such film.
    >
    > Also, all the above applies only to the middle of the response curve.
    > There are usually "shoulder" and "toe" regions with lower contrast for
    > the shadows and extreme highlights respectively. And the centre section
    > really isn't exactly straight either.


    Thanks for the reply. Is this then the reason that images from digital
    cameras look different from images from film cameras ? I get the
    impression that digital photos better reproduce reality, while film is a
    bit surreal.
    --

    Alfred Molon
    ------------------------------
    http://groups.yahoo.com/group/Olympus_405080/
    Olympus 5060 resource - http://myolympus.org/5060/
    Olympus 8080 resource - http://myolympus.org/8080/
     
    Alfred Molon, Oct 7, 2004
    #7
  8. Alfred Molon

    Alfred Molon Guest

    Alfred Molon, Oct 7, 2004
    #8
  9. Alfred Molon wrote:
    []
    > Thanks for the reply. Is this then the reason that images from digital
    > cameras look different from images from film cameras ? I get the
    > impression that digital photos better reproduce reality, while film
    > is a bit surreal.


    Yes, and the way that white level "clips" quite sharply - anything over
    white is lost whereas it can be recovered on film.

    Having worked with television (and basically linear sensors with gamma
    correction) for about 40 years I much prefer the "digital" to the "film"
    look, although I could imagine a cinema-goer might prefer the opposite!

    Cheers,
    David
     
    David J Taylor, Oct 7, 2004
    #9
  10. Alfred Molon wrote:
    ....
    >
    > Thanks for the reply. Is this then the reason that images from digital
    > cameras look different from images from film cameras ? I get the
    > impression that digital photos better reproduce reality, while film is a
    > bit surreal.


    It is a reason not the reason. There are many differences, but I would
    guess that it is a primary reason.


    --
    Joseph E. Meehan

    26 + 6 = 1 It's Irish Math
     
    Joseph Meehan, Oct 7, 2004
    #10
  11. Alfred Molon <> writes:

    >Thanks for the reply. Is this then the reason that images from digital
    >cameras look different from images from film cameras ? I get the
    >impression that digital photos better reproduce reality, while film is a
    >bit surreal.


    I don't think so. The response of the film is only one part of the
    imaging process. If you're making prints, the film with gamma -0.6 is
    printed onto paper with gamma of about -1.7, giving a gamma for the
    whole system which is the product of the two values - about 1.0. So the
    negative/print process, as a whole, is close to linear reproduction.

    Meanwhile, CCD *sensors* are linear, but the data from them immediately
    undergoes nonlinear processing in the camera (unless you shoot raw).
    JPEG images have gamma-corrected data in them, so in a sense the overall
    camera gamma is 0.45 for digital cameras. Highlights are probably
    compressed too, in a simulation of film's "shoulder" response.

    I think the more important differences are the shape of the MTF
    response, and the handling of over and under exposure.

    The MTF curve of film and lenses both tends to have 100% contrast at
    very low frequencies, dropping slowly and gradually to nothing at some
    fairly high spatial frequency. The curve looks sort of like the
    profile of a sandy beach. A digital camera using the same size sensor
    and the same lens will tend to have a response that stays near 100% for
    a while instead of dropping gradually, but when it gets to about 70% of
    the Nyquist frequency it will begin to drop very rapidly (partially due
    to the antialiasing filter). The curve looks more like the profile of
    a cliff. With current digital cameras, the point at which the MTF
    drops to zero (resolution limit) may be half that of the film.

    For a hypothetical example, assume that the film resolution limit is 100
    lp/mm, while the digital camera's limit is 50 lp/mm. Because of the
    difference in MTF shape, the digital camera will have *higher contrast*
    than the film system at 10, 20, 30, and probably 40 lp/mm, but the
    digital can't resolve anything useful at all higher than 50 lp/mm. The
    film image continues to resolve detail, at lower contrast, out to double
    the frequency. If you view prints from both systems at normal viewing
    distance, the digital one will look better because of the higher
    contrast at frequencies that are highly visible to your eyes. But if
    you get close to the prints, the film one will have subtle fine details
    that just aren't there in the digital image. The film has a "softer"
    and more gradual resolution limit.

    As digital camera resolution continues to improve, while film changes
    much more slowly, these differences will get less pronounced. If we
    ever get digital cameras with so many pixels that the lens determines
    the resolution limit, not the sensor, the MTF will end up looking a lot
    like current film systems. (And people will complain that the "per
    pixel" resolution of the camera is poor - that the images are horribly
    fuzzy at 100%).

    The other thing that film does well is over- and under-exposure
    handling. The shoulder and toe of film mean that, for several stops
    above and below the "linear range", film continues to capture some
    information with reduced contrast. That means that deep shadows and
    strong highlights still contain *some* image information, rather than
    being a featureless black hole or white hole in the image respectively.

    CCDs are pretty linear up to the point where the wells overflow, and
    then the electrons are lost (or spill into adjacent pixels, which is
    worse). This means that a CCD sensor, itself, has a very hard white
    clipping behaviour. If you want to build a CCD camera system that
    simulates a shoulder, you need to expose the CCD so that the "linear"
    range is nowhere near CCD saturation, and use most of the actual CCD and
    A/D converter range for the nonlinear shoulder. You can only afford
    this if you have lots of dynamic range to start with. So, in practice,
    both video and still digital cameras tend to have unpleasant white
    clipping in circumstances where film, particularly negative film, would
    cope better.

    Another effect: film grain noise doesn't look like digital sensor
    noise.

    Dave
     
    Dave Martindale, Oct 7, 2004
    #11
  12. "Jeremy Nixon" <> wrote in message
    news:...
    > Alfred Molon <> wrote:
    >
    >> I got this impression when processing the scans of my brother's slides.
    >> Very strange colours and brightness levels. While CCDs are linear
    >> devices, where the output (before gamma) is a linear function of the
    >> light level, I get the impression that film does not perform like that
    >> and that the response curve is far from linear. Is this the case?

    >
    > Not linear at all. Kodak used to publish little pamphlets with all the
    > information about each of their types of film; they probably still do.
    > One of the things you get there is the response curve, which is far from
    > linear, even taking gamma into account.
    >
    > --
    > Jeremy |


    Those who used to use film for quantitative intensity measurements developed
    functions that would linearize the response over most of the response curve.
    No one function worked for all emulsions.
     
    Marvin Margoshes, Oct 7, 2004
    #12
  13. "Jim" <> wrote in message
    news:zZ19d.2021$...
    >
    > "Alfred Molon" <> wrote in message
    > news:...
    >> I got this impression when processing the scans of my brother's slides.
    >> Very strange colours and brightness levels. While CCDs are linear
    >> devices, where the output (before gamma) is a linear function of the
    >> light level, I get the impression that film does not perform like that
    >> and that the response curve is far from linear. Is this the case?
    >> --

    > No film is not linear. Reciprocity failure is just another description of
    > non-linearity.
    > Jim
    >

    Not correct. Sorry about that. In photography, exposure is a linear
    product of intensity and time. If the intensity is 10 times higher,
    reducing the exposure time to one-tenth gives the same exposure and the same
    reponnse by the emulsion; that relation is called reciprocity. A silver
    halide grain in a photographic emuslion must absorb two photons to become
    subject to conversion to a silver particle in the development process. If
    one photon is absorbed, and a second one is not absorbed in a time that
    depends on the type of emulsion, the particle reverts to its initial state.
    For ythat reason, with very long exposures an emulsion is subject to
    reciprocity failure.

    Once two photons are absorbed in a short enough time, absorbing a third
    makes no difference.
     
    Marvin Margoshes, Oct 7, 2004
    #13
  14. Re: Film is not linear ? It's partly linear.

    Alfred Molon wrote:
    > I got this impression when processing the scans of my brother's slides.
    > Very strange colours and brightness levels. While CCDs are linear
    > devices, where the output (before gamma) is a linear function of the
    > light level, I get the impression that film does not perform like that
    > and that the response curve is far from linear. Is this the case?


    Yes, CCDs can be extremely linear response devices if operated carefuly,
    even over a dynamic range approaching 100,000. That's one reason why
    astronomers have all but abandoned film. The response of film is
    non-linear and is described by what is called the "characteristic curve"
    or "H and D curve." The vertical axis is the amount of exposure
    (darkening or density) and the horizontal axis is the amount of light
    exposure. This curve looks like a long drawn-out "S" or integral sign,
    if you are familiar with calculus. The bottom part is called the toe,
    when film is not very sensitive to light as it just begins to darken.
    The middle part is quite straight and is called the linear part. The
    top part turns over because the film begins to become fully exposed. In
    scientic applications, film could be exposed to spots of light with a
    range of intensities so the the curve could be established for that
    particular emulsion. Then all densities could be converted to
    intensities. It works quite well except near the ends of the curve. I
    use to "pre-flash" film for long exposures (hours) to make it slightly
    exposed so that I started the real exposure on the linear part.
     
    Joseph Miller, Oct 7, 2004
    #14
  15. Alfred Molon

    Alfred Molon Guest

    In article <ck3nh0$a6$>, says...

    > As digital camera resolution continues to improve, while film changes
    > much more slowly, these differences will get less pronounced. If we
    > ever get digital cameras with so many pixels that the lens determines
    > the resolution limit, not the sensor, the MTF will end up looking a lot
    > like current film systems. (And people will complain that the "per
    > pixel" resolution of the camera is poor - that the images are horribly
    > fuzzy at 100%).


    I thought we had already reached that point. Not many lenses are capable
    of resolving 2.8 micrometer pixels - that's 180 lp/mm.
    --

    Alfred Molon
    ------------------------------
    http://groups.yahoo.com/group/Olympus_405080/
    Olympus 5060 resource - http://myolympus.org/5060/
    Olympus 8080 resource - http://myolympus.org/8080/
     
    Alfred Molon, Oct 7, 2004
    #15
  16. Alfred Molon

    Guest

    Kibo informs me that Alfred Molon <> stated
    that:

    >I got this impression when processing the scans of my brother's slides.
    >Very strange colours and brightness levels. While CCDs are linear
    >devices, where the output (before gamma) is a linear function of the
    >light level, I get the impression that film does not perform like that
    >and that the response curve is far from linear. Is this the case?


    Yes, that's correct.

    --
    W
    . | ,. w , "Some people are alive only because
    \|/ \|/ it is illegal to kill them." Perna condita delenda est
    ---^----^---------------------------------------------------------------
     
    , Oct 8, 2004
    #16
  17. >
    > Not correct. Sorry about that. In photography, exposure is a linear
    > product of intensity and time. If the intensity is 10 times higher,
    > reducing the exposure time to one-tenth gives the same exposure and the same
    > reponnse by the emulsion; that relation is called reciprocity. A silver
    > halide grain in a photographic emuslion must absorb two photons to become
    > subject to conversion to a silver particle in the development process. If
    > one photon is absorbed, and a second one is not absorbed in a time that
    > depends on the type of emulsion, the particle reverts to its initial state.
    > For ythat reason, with very long exposures an emulsion is subject to
    > reciprocity failure.
    >
    > Once two photons are absorbed in a short enough time, absorbing a third
    > makes no difference.
    >
    >


    Hmm, I'm sceptical to say the least about this extreme
    oversimplification. The creation of the latent image in film emulsion is
    a chemical reaction and as such it is far from being precise.

    Anyway all of this theory is not really relevant. The fact is that with
    very long, or very short exposures film stops behaving in a linear
    fashion. In between it's close to linear, but not precisely.

    The point at which this occurs is also not precise. Like all things in
    chemistry it's a gradual change. That's why the response curve is not
    exactly straight.

    Perfectly straight response curves can only happen in the digital world,
    not the real world. Silver halide activation is a real world process.
     
    Eugene O'Brien, Oct 8, 2004
    #17
  18. >
    > Not correct. Sorry about that. In photography, exposure is a linear
    > product of intensity and time. If the intensity is 10 times higher,
    > reducing the exposure time to one-tenth gives the same exposure and the same
    > reponnse by the emulsion; that relation is called reciprocity. A silver
    > halide grain in a photographic emuslion must absorb two photons to become
    > subject to conversion to a silver particle in the development process. If
    > one photon is absorbed, and a second one is not absorbed in a time that
    > depends on the type of emulsion, the particle reverts to its initial state.
    > For ythat reason, with very long exposures an emulsion is subject to
    > reciprocity failure.
    >
    > Once two photons are absorbed in a short enough time, absorbing a third
    > makes no difference.
    >
    >


    You seem to be contradicting yourself. In your previous post you said
    that functions had to be created to try to linearise most of the
    response curve, and now you seem to be saying that the response curve is
    already linear?
     
    Eugene O'Brien, Oct 8, 2004
    #18
  19. >
    > Thanks for the reply. Is this then the reason that images from digital
    > cameras look different from images from film cameras ? I get the
    > impression that digital photos better reproduce reality, while film is a
    > bit surreal.


    This one is a bit dubious. The fact is that you have finite storage
    space for the image data, so digital has to make approximations. It has
    to approximate intensity levels to 12 bits (for most CCD's I think) of
    data for each colour. Colour in reality is not 8bit, or 12bit, it's not
    even 16bit. It's infinite. You cannot say that there are 16.8 million
    colours in the world.

    Digital these days may more closely represent colours the way the human
    eye sees them, but in it's operation it's entirely artificial. Clipping
    for example is a purely digital concept, it doesn't happen in reality,
    or on film.

    A pure saturated white in a 8bit per pixel image is represented by rgb
    values (255,255,255). In reality there is no such thing as a pure
    saturated white. There's white like a white car on a sunny day, or white
    like the blast of a strobe, or white like the sun at midday in summer. A
    digital camera will likely represent them all as having exactly the same
    colour value. Whereas clearly to the human eye they're very different.
     
    Eugene O'Brien, Oct 8, 2004
    #19
  20. Alfred Molon <> writes:

    >> As digital camera resolution continues to improve, while film changes
    >> much more slowly, these differences will get less pronounced. If we
    >> ever get digital cameras with so many pixels that the lens determines
    >> the resolution limit, not the sensor, the MTF will end up looking a lot
    >> like current film systems. (And people will complain that the "per
    >> pixel" resolution of the camera is poor - that the images are horribly
    >> fuzzy at 100%).


    >I thought we had already reached that point. Not many lenses are capable
    >of resolving 2.8 micrometer pixels - that's 180 lp/mm.


    The cameras that use sensors with 2.8 um pixels are the ones with small
    sensors and very short focal length lenses - and they do have that sort
    of resolution. It's not a surprise either.

    Suppose you have a sensor that is 1/5 the dimensions of a 35 mm film
    frame. A "normal" lens for this camera will have a focal length of
    about 10 mm. If you start with the design for a 50 mm f/2 normal lens
    for a 35 mm camera and simply scale it down by a factor of 5 (divide all
    the radii, thicknesses, diameters, and spacings by 5), you will get a
    good 10 mm f/2 lens suitable for your camera. The new lens will have
    the same angular covering power as the original 50 mm lens, so it will
    cover the sensor nicely.

    And the aberrations *also* scale down by a factor of 5. So, considering
    aberration blur only, if the 50 mm original lens would resolve 50 lp/mm,
    the new 10 mm lens will resolve 250 lp/mm. This is what you get simply
    scaling an existing design; nothing exotic at all.

    (Now, it's not quite so rosy in reality, because diffraction blur does
    *not* scale down, and the new lens has as much diffraction blur at f/4
    as the original lens had at f/20, referred to equal-sized prints. So
    you can't stop the 10 mm lens down as far).

    So the cameras that have 2.8 um pixel pitch also have short focal length
    high-spatial-resolution lenses. Few SLR lenses will resolve 180 lp/mm,
    but even the highest-resolution DSLRs have pixel pitches of 6-7 um not
    2.8. This corresponds to about 70 lp/mm, which is getting above what
    *some* SLR lenses deliver, but certainly not all.

    Because of this resolution scaling effect, at 4 or 6 megapixels, the
    sensor pixel pitch is the thing that sets the resolution of the camera,
    as long as the lens is fairly decent, and it doesn't matter whether
    we're talking about a small-sensor P&S camera or a dSLR.

    On the other hand, if we ever get 3 um pixel pitch in a full-frame DSLR
    sensor, *then* the lens will usually be the resolution limit. That's
    what I was talking about. (But that's 96 MP per image; it will take a
    while yet to get there).

    Dave
     
    Dave Martindale, Oct 8, 2004
    #20
    1. Advertising

Want to reply to this thread or ask your own question?

It takes just 2 minutes to sign up (and it's free!). Just click the sign up button to choose a username and then you can ask your own questions on the forum.
Similar Threads
  1. Dave

    Linear vs. Circular Polarizer ?

    Dave, Dec 9, 2003, in forum: Digital Photography
    Replies:
    13
    Views:
    6,454
    Bob Salomon
    Dec 10, 2003
  2. Neal Matthis

    What is Linear Processing

    Neal Matthis, Jan 29, 2004, in forum: Digital Photography
    Replies:
    8
    Views:
    4,755
  3. Jeff Durham

    Polarizing filter -- circular or linear?

    Jeff Durham, Jun 14, 2004, in forum: Digital Photography
    Replies:
    13
    Views:
    3,457
    Dave Martindale
    Jun 15, 2004
  4. Destin_FL

    Linear or circular polarizer for digital

    Destin_FL, Jul 11, 2004, in forum: Digital Photography
    Replies:
    5
    Views:
    4,189
    Bob Salomon
    Jul 12, 2004
  5. Nick Lamendola

    Circular vs Linear Polarizer - How do you tell them apart?

    Nick Lamendola, Jul 24, 2004, in forum: Digital Photography
    Replies:
    4
    Views:
    1,589
    Alan D
    Jul 25, 2004
Loading...

Share This Page