D70 Custom Curves and Noise

Discussion in 'Digital Photography' started by Xavier, Jul 23, 2004.

  1. Xavier

    Xavier Guest

    I have been reading more about the custom curves function on Nikon's
    DSLRs and have this question:

    Are the custom curves applied as part of the exposure process or are
    they applied after exposure?

    I am just trying to understand if applying customs curves in camera to
    improve shadow detail will result in less shadow noise as oppose to
    when the curves are applied post-capture in Nikon Capture or other
    post-processing software to achieve the same effect. I understand that
    applying curves are applied on a log-scale in camera and linearly in
    the post-processing software. However, if the difference in applying
    the curves in camera or in post-processing is merely in the scale in
    which they are applied, then you won't get any noise improvement when
    shadow details are lifted. But if applying the custom curves in camera
    invovles changing the gain of the sensor amplifiers etc, then applying
    the curves in camera will definitely result in lower noise in the
    shadows. Can't figure out which is the case.

    Ideas anyone?
     
    Xavier, Jul 23, 2004
    #1
    1. Advertisements

  2. Xavier

    Jeremy Nixon Guest

    Xavier <> wrote:

    > Are the custom curves applied as part of the exposure process or are
    > they applied after exposure?


    The curve is used when converting the linear RAW data into a picture.

    --
    Jeremy |
     
    Jeremy Nixon, Jul 23, 2004
    #2
    1. Advertisements

  3. Xavier

    Guest

    (Xavier) wrote:

    > Are the custom curves applied as part of the exposure process or are
    > they applied after exposure?


    Everything before the ADC's are as linear as possible. Raw samples
    from the sensor are linear samples. This is a good thing.

    > But if applying the custom curves in camera
    > invovles changing the gain of the sensor amplifiers etc, then applying
    > the curves in camera will definitely result in lower noise in the
    > shadows.


    You can't "lower noise" in a single pixel! Equivalence,
    www.google.com: "second law of thermodynamics". This is basic signal
    processing. Pixel SNR is direct related to pixel value (ie, brighter
    pixels have a better SNR). www.google.com: pixel noise CCD poisson.
    Numerous relevant hits. Example:

    http://www.fbw.hs-bremen.de/~olbers/sternw/sonne/doc/noise.htm

    Changing the gain of the pre-ADC amplifiers -- even nonlinearly --
    doesn't change the underlying SNR of a pixel. Your only option is to
    draw a larger sample of photons, examples:

    a) increase exposure (dynamic range problems can interfere)
    b) group pixels within the image (basis of noise reduction algorithms,
    high frequency rolloff can be bothersome)
    c) group pixels between images (stacking, impractical for moving
    scenes)
    d) various combinations of the above (use (a) and (c) to work around
    potential dynamic range problems of 'a', but you still are limited to
    static scenes)

    [Other sources of noise exist (amplifier noise, thermal electrons in
    the sensor, etc), but the Poisson noise dominates.]
     
    , Jul 24, 2004
    #3
    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.