Can anyone verify this 10D anomaly?

Discussion in 'Digital Cameras' started by Roland Karlsson, Sep 2, 2004.

  1. Roger N Clarke found in a very interesting test
    that the 10D clipped one stop lower when saving
    as JPEG compared to saving as RAW. This will
    zap all the white parts in the RAW original.

    http://clarkvision.com/imagedetail/dynamicrange

    This is a rather peculiar behaviour, not easily
    to motivate. Now, Roger may be wrong of course.

    Is it possible that someone can make the following
    test. Take a picture, with a small white object, using
    several manual exposures, both with RAW and JPEG.

    Then, comparing the JPEG to a 16 bit TIF (extracted
    from the RAW) to see whether the small white object
    contains details at the same exposures or the white
    object loses the details at a lesser exposure.


    /Roland
     
    Roland Karlsson, Sep 2, 2004
    #1
    1. Advertisements

  2. Roland Karlsson

    Ryadia Guest

    ------------------
    Quite apart from this technical explanation... I have found from day one
    that shooting in .jpg mode when there are bright highlights will result
    in the complete loss of detail in those highlights.

    I originally put it down to multi-spot metering but changing that to
    centre spot only moderately altered the effect. I still shoot .jpg mode
    indoors, under flash with great success but no more will I shoot in
    sunlight any other method than RAW.

    oddly enough, the small preview image often shows the detail missing
    later in the .jpg file!

    Ryadia.
     
    Ryadia, Sep 3, 2004
    #2
    1. Advertisements

  3. Roland Karlsson

    JPS Guest

    In message <>,
    Dynamic range at the highlight end has nothing to do with 8-bit vs 16
    bit. That's just a matter of level granularity. On the dark end, the
    higher bit depth will allow more distinguishable shadows. It has
    nothing to do with TIFF vs JPEG, either, as the linear TIFF is simply a
    TIFF that has no gamma correction. A 16 bit TIFF can be gamma-corrected
    to, as are the TIFFs made by any RAW converter, and they may have the
    same clipping point as a camera JPEG.
    --
     
    JPS, Sep 3, 2004
    #3
  4. Roland Karlsson

    JPS Guest

    In message <>,
    That is impossible. You are viewing the JPEG in the review. The 10D
    never renders from RAW to the screen; it renders to JPEG and then to the
    screen.
    --
     
    JPS, Sep 3, 2004
    #4
  5. Hi guys.
    The test I did was raw plus jpeg on the same image for the
    canon 10D. The jpeg is clipped and shows less highlights
    than the raw (16-bit tiff converted). This is seen in Figure 4
    on the page: the highlights on the dog are clipped.

    I often see this problem with the 10D and 8-bit jpegs
    versus 16-bit raw (really 12 bit). In particular
    I have had this problem a lot imaging white birds, and often
    underexpose 1 to 2 stops to prevent clipping. I like to avoid
    raw because I can get so many more images on a card.
    My new 1D Mark II seems to meter better in this regard, but the
    problem is still there in the jpegs versus raw. The 1DII seems
    to meter better if there is a small bright area in the scene
    and adjusts exposures better.

    Roger
     
    Roger N. Clark (change username to rnclark), Sep 3, 2004
    #5
  6. Roland Karlsson

    Ryadia Guest

    The preview image I write about is the preview in the .jpg file, not the one
    you refer to.

    Ryadia
    ----------------------
     
    Ryadia, Sep 3, 2004
    #6
  7. Roland Karlsson

    JPS Guest

    In message <>,
    I now have absolutely no idea what you're talking about.

    --
     
    JPS, Sep 3, 2004
    #7
  8. Roland Karlsson

    Don F Guest

    If I may interrupt, I think that Ryadia's comments are based on
    conclusions he has made based on shooting in .JPG mode not RAW mode.
    Ryadia wrote:
    Quote:
    Quite apart from this technical explanation... I have found from day one
    that shooting in .jpg mode when there are bright highlights will result
    in the complete loss of detail in those highlights.
    End Quote:
    Of course there could be a display adjustment based on what jpeg level is
    chosen.

    Don F
     
    Don F, Sep 3, 2004
    #8
  9. Roland Karlsson

    Don F Guest

    --------------
    Are you saying that the 10D display is essentially useless in highlight
    evaluation?
    The D70 preview display clearly shows blown highlights while shooting in
    RAW mode. The D70 display will also select the highlight area with a dotted
    line for zoomed in evaluation.
    Do all cameras convert RAW to JPG for display purposes? If so, what is
    Nikon doing differently?
    I know this thread references the 10D specifically and my comment is not
    meant as a criticism. I am still low on the digital photography learning
    curve so I want to understand any comments made by knowledgeable posters.
    Don F
     
    Don F, Sep 3, 2004
    #9
  10. Personally I consider this to be an anomaly. The ISO sensitivity
    for digital sensors is defined by the white clipping point. If
    you lower the clipping point you increase the ISO sensitivity.
    Therefore, the JPEG picture has twice the ISO sensitivity as
    the TIF (converted from RAW) picture.


    /Roland
     
    Roland Karlsson, Sep 3, 2004
    #10
  11. SNIP
    Not necessarily. It is only a part of the ISO speed rating decision,
    and another part is based on the noise level.

    In particular, a quote from a Working Draft for the ISO 12232
    standard:
    <quote>
    With appropriate electrical or digital gain, a DSC can provide an
    appropriate output signal level for a range of sensor exposure levels.
    The maximum exposure level is the exposure level where typical picture
    highlights will be clipped as a result of saturating the image sensor
    signal capacity or reaching the camera signal processing maximum
    signal level. The minimum exposure level depends on the amount of
    noise that can be tolerated in the image. These situations lead to two
    different types of speed values, saturation signal based values, and
    noise based values. The ISO speed is preferably determined using a
    noise-based method. The saturation-based value is preferably used to
    indicate the camera's overexposure speed latitude. A second
    noise-based value is preferably used to indicate the camera's
    underexposure speed latitude. For some types of DSCs, such as those
    employing lossy compression methods that significantly affect measued
    image noise values and cannot be bypassed, it is not possible to
    correctly determine the noise based ISO speed. In such cases, the ISO
    speed of the camera is determined using the saturation-based
    measurement, and the ISO speed latitude values are not reported. In
    other cases, the noise-based ISO speed may be lower than the
    saturation-based speed, in which case the saturation based-speed is
    reported.
    </quote>

    In other words, if image quality is of major importance (e.g.
    photographic studio with controlled lighting), a saturation based
    rating is preferred. However, if camera shutter speed or small
    apertures are more important, then a noise based rating is more
    common, unless it is compromised by unavoidable(!) lossy compression,
    or when it turns out to be lower than the other rating.

    Bart
     
    Bart van der Wolf, Sep 3, 2004
    #11
  12. Hmmm .. wise words indeed in this draft.

    The experments made by Roger N Clark hints at 10D using saturation
    based ISO for RAW and using faulty saturation based ISO for JPEG.


    /Roland
     
    Roland Karlsson, Sep 3, 2004
    #12
  13. Roland Karlsson

    JPS Guest

    In message <>,
    What, exactly, do you mean by that?
    --
     
    JPS, Sep 3, 2004
    #13
  14. Roland Karlsson

    JPS Guest

    In message <[email protected]>,
    It's not as good as it could be. It could show something blown out in
    the JPEG, that isn't actually blown out at all in the RAW data.
    So does the 10D. It shows them for the JPEG settings, though, and not
    for the RAW file. It is monochrome, too, and can actually miss clipping
    in the red and blue channels. The monochrome histogram is unevenly
    weighted from the separate R, G, and B histograms; 60% green, 30% red,
    and 10% blue is the typical weighting. This means that if the blue were
    500, and the green and red were 150 each, the weighted total would be
    150*0.6 + 150*0.3 + 500*0.1 = 185. This would not clip the histogram,
    but the blue channel would be severely clipped.
    No camera shows RAW, AFAIK. RAW does not look very good unless it is
    converted to gamma-corrected RGB space.
    I don't know what Nikon is doing.
    --
     
    JPS, Sep 4, 2004
    #14
  15. Roland Karlsson

    JPS Guest

    In message <>,
    What I remember is that the update rolled off the extreme shadows faster
    with a setting of 0; as if the old setting of 0 was equivalent to the
    new setting of 10, or, the shadow roll-off became less or none by
    default.

    I don't think it actually clips, though. I think it just adjusts a
    curve.
    --
     
    JPS, Sep 4, 2004
    #15
  16. Roland Karlsson

    JPS Guest

    In message <>,
    Yes, I just verified that it does clip. It also changes the gamma so
    that 255 is still 255.
    --
     
    JPS, Sep 6, 2004
    #16
    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.