Nikon D2X lowest high ISO noise?

Discussion in 'Nikon' started by Rich, Nov 22, 2005.

  1. Rich

    Rich Guest

    This is in the most recent issue of "Popular Photography."
    At high ISOs it achieves the lowest noise of any DSLR up to 1600 ISO,
    without using a "blurring filter." Seems to contradict other things
    I've read. In any case, it's sensor is a Sony CMOS.
    Rich, Nov 22, 2005
    1. Advertisements

  2. But the in camera processing of the image averages adjacent
    pixels to reduce noise, thus loweering spatial resolution.
    You can see this on nikon cameras on the test images on

    Roger N. Clark (change username to rnclark), Nov 23, 2005
    1. Advertisements

  3. Rich

    Rich Guest

    Does that apply to RAW files from the Nikon?
    Rich, Nov 23, 2005
  4. I do not know. Are the test images on dpreview from raw?
    Roger N. Clark (change username to rnclark), Nov 23, 2005
  5. Rich

    Jeremy Nixon Guest

    There is no noise reduction on the raw files.

    "Test" images on dpreview are typically JPEGs shot at some arbitrary set
    of camera settings, and are therefore meaningless.
    Jeremy Nixon, Nov 23, 2005
  6. Rich

    Rich Guest

    Highest ISO however in the sample images was 400.
    The test images go to 3200 and are in JPEG
    but if you look at the noise graph, the Nikon is lowest at
    Rich, Nov 23, 2005
  7. Rich

    JPS Guest

    In message <>,
    Why doesn't someone with web space start a library of RAW sample from
    various cameras, like blackframes, totally clipped images, Color-checker
    charts at different levels of exposure, etc. Of course, this would take
    a lot of storage.

    You can tell so much about how a camera works from RAW data. You can
    take a blackframe, make a histogram of it and measure the deviation, the
    extremes. You can determine if blackpointing is done before writing the
    file. You can do horizontal and vertical blurs to look for banding
    artifacts, separate these from the random elements of noise; maybe come
    up with solutions. RAW tells so much, yet everyone is always talking
    about and demonstrating converted files (sometimes with heavy JPEG

    Part of the reason, I think, for the slow progress with RAW conversion
    has to do with the fact that very few people are examining real RAW
    data, so only the developers are generally looking at it, and probably
    without a lot of criticism. Same goes for the manufacturers. How hard
    would it be for Canon to take the horizontal and vertical blackpoints
    for each line, and apply them before writing a RAW file, and recording
    the changes as metadata? They feel no pressure to do so, because no one
    is talking knowledgeably about the problems, except a few people. They
    are relying on the other users who defend the products with "what do you
    expect when you under-expose?" to squelch complaints.
    JPS, Nov 23, 2005
  8. But the in camera processing of the image averages adjacent
    RAW files don't have pixels in the normal sense. So they can only be
    viewed after some processing. _If_ you can define the parameters for
    this processing, you can reduce spacial resolution to improve the noise.

    Michael Schnell, Nov 23, 2005
  9. Rich

    JPS Guest

    In message <>,
    Not really. You can view them in a number of ways; as a greyscale
    (which will have a checkerboard pattern in any areas where the three
    channels are not nearly equal in intensity). You can then split the 4
    channels (1 red, 2 greens, and 1 blue) so that you have 4 separate
    images in greyscale representing the channels.

    You can also view RAW data with the pixels colored based on the RAW
    channel that they came from. You can also view an full RGB image that
    is made by interpolating the missing samples in each color plane.
    The noise could most easily be addressed in the RAW state, where it has
    a much higher frequency component (at the nyquist) and is therefore more
    obvious. The pixels from different color channels are at different
    levels, typically, so it is a little different dealing with the data,
    but you could normalize the levels based on local luminance, or, you
    could just filter each color channel, but then you no longer have noise
    at the nyquist of the image, so the ratio of AA-limited frequencies to
    the highest frequency of noise is higher, so signal and high-frequency
    noise are less distinct.
    JPS, Nov 24, 2005
    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.