How much information does jpeg lose?

Discussion in 'Digital Cameras' started by Al Dykes, Apr 15, 2007.

  1. Al Dykes

    Al Dykes Guest

    If I shoot RAW and then in Photoshop, export it to a jpeg
    with the highest ssetting, how much information is lost?

    If I then bring that jpeg into PS and save it as a PSD, how much have
    I lost?
     
    Al Dykes, Apr 15, 2007
    #1
    1. Advertisements

  2. Al Dykes

    John Bean Guest

    Impossible to quantify. Depends on the contents of the
    image.
    None.
     
    John Bean, Apr 15, 2007
    #2
    1. Advertisements

  3. You can do an amazing amount of adjustments to the RAW image itself in
    recent versions of ACR and the release version of Lightroom. Using one
    of those, you could quickly go back to the source image and start
    afresh, or use the full image already adjusted to generate a new TIFF or
    PSD.
     
    John McWilliams, Apr 15, 2007
    #3
  4. Al Dykes

    John Bean Guest

    Of course not. But that wasn't the question... or was it?

    On reflection perhaps that's what the OP was getting at. In
    that case well spotted ;-)
     
    John Bean, Apr 15, 2007
    #4
  5. Al Dykes

    Al Dykes Guest

    What is "ACR". The initials escape me. "A" probably stands for Adobe
    but I need to buy another clue.
     
    Al Dykes, Apr 15, 2007
    #5
  6. Al Dykes

    Al Dykes Guest


    I know that you can't replace lost info and that saving a jpeg on
    itself is a Bad Idea.

    Thanks
     
    Al Dykes, Apr 15, 2007
    #6
  7. Adobe Camera Raw. The raw converter in PS. PSE and other Adobe products.
     
    Ed Ruf (REPLY to E-MAIL IN SIG!), Apr 15, 2007
    #7
  8. Al Dykes

    Al Dykes Guest


    That's it. I use RAW in PS all the time. The initials drew a blank this morning..
     
    Al Dykes, Apr 15, 2007
    #8
  9. Al Dykes

    HEMI-Powered Guest

    Today, Al Dykes made these interesting comments ...
    if a tree falls in a forest and nobody is there to see/hear it,
    does it make any noise?
     
    HEMI-Powered, Apr 15, 2007
    #9
  10. Al Dykes

    John Bean Guest

    It didn't fall.
     
    John Bean, Apr 15, 2007
    #10
  11. Al Dykes

    ben brugman Guest

    This filosofical question has been answered by scientist. As science stands
    at this moment the anwser is YES.
    And allthough observers ALWAYS influence an experient, taking away the
    observers does not take away the sound, but as with Schrodinger's cat one
    can not say that the tree has fallen or not fallen until an observation is
    made. So the tree might still be standing.

    http://en.wikipedia.org/wiki/Schrödinger's_cat

    For the question of how much information is lost, one first has to anwser
    how much information was there in the scene, how much information was there
    in the 'raw format' and what of that information is lost. The amount of
    information is at most equal to the amount of data, but often it is far
    less, but it's very difficult to express how much information there is in a
    picture.
    Reducing a picture to only exactly 9 colors does reduce the data a lot in a
    picture, and hardly any information is lost in the picture. (At least if the
    consumers of the picture are humans). If the picture is used for coding
    something then a lot of information is lost.

    The coordinates (3) of a picture, the angles (3), the time, the
    shutterspeet, focal length, focusdistance, aperature and format information
    determine a picture completely. This is only a few bytes of information and
    one only needs a time/travel machine to completely reconstruct the picture
    from this information.

    ben
     
    ben brugman, Apr 15, 2007
    #11
  12. Al Dykes

    ASAAR Guest

    According to the NYTimes, it loses all the information that's not
    fit to print. :)
     
    ASAAR, Apr 15, 2007
    #12
  13. Al Dykes

    ray Guest

    Compared to the raw file, you've at least lost all the extra dynamic range
    - most raw files use 12bits of data per channel - jpeg saves 8bits.
     
    ray, Apr 15, 2007
    #13
  14. Al Dykes

    ben brugman Guest

    No this is not completely true, because most of the 12 bits raw are
    compressed
    in the jpeg 8 bit format, this is done by applying a curve. So raw does not
    have
    2 bits on both sides of the 8 bit jpg with exposure latitude. So most of the
    dynamic
    range is kept. The detail of this dynamic range is less. And on both sides a
    bit of
    dynamic range is lost.

    Lot's of camera's not having raw have more than 8 bits internal resolution,
    this would
    be useless if there wouldn't be a 'compression' like a curve.

    So raw files do not have 4 bit extra of dynamic range, but depending on the
    settings
    somewhere from 0 to 2 bit of extra dynamic range. (High contrast setting for
    jpeg
    does do extra compression).

    Then the scene which you started with may have less then 12 (reals) bits of
    dynamic range.
    So then you don't loose either.

    But raw gives you definitely more control over your picture in
    postprocessing (on the pc),
    for jpeg you have to make a lot of decisions on the processing before you
    press the
    button. (Contrast, saturation, white balance, sharpness etc.). With raw you
    can postpone
    these decisions and see what the effect of each decision is.

    Ben.
     
    ben brugman, Apr 15, 2007
    #14
  15. Al Dykes

    John Sheehy Guest

    This is a somewhat different issue than shooting JPEG in the camera, if
    you are tweaking your conversions. Shooting JPEGs in the camera (or
    accepting defaults in a converter) lose quite a bit of RAW highlights to
    clipping. Depending on the white balance, saturation, and subject
    colors, this can literally be on the order of 4 to 5 stops (red channel
    in red flowers in deep blue sky shade).

    You lose ranges of highlights, the extents of which depend on your
    exposure and contrast settings in the converter, the white balance, and
    the saturation settings (high saturation clips more reds and greens with
    certain colors - mainly reds, oranges, and yellows). If everything you
    need remains in the JPEG as far as highlight ranges are concerned,
    though, then you aren't missing anything up there - you can only lose it
    if its there. Most RAW files are quite poorly exposed in low- and
    medium-contrast scenes with default exposure compensation. Of course, if
    you know that the extra highlights are there, you can use them to your
    advantage,

    If you're happy with the recording of the highlights, then the issues are
    fewer. JPEG compression puts little squares all over the image, and if
    you do any post processing, they have a greater chance of becoming
    visible, like increasing the contrast of midtones or highlights, or
    boosting the shadow areas. Reduction to 8-bits gamma-adjusted data
    posterizes the highlight areas as well, giving them the same PP issues.
    Nothing further. PSD can convey everything in a JPEG properly.

    If you're thinking about using JPEG as a more efficient way to store your
    images, I would recommend agaisnt it, unless you are happy that you will
    never need to PP them further. JPEG should be the last stop for an
    image, if you are interested in maximum quality and/or flexibility. 8-
    bit too. In the case of a camera like the Leica M8 that saves 8-bit
    gamma-adjusted RAWs (DNGs, specifically), the only compromise is the
    posterization of the highlight areas, but they contain none of the JPEG
    issues other than bit-depth. I would do all PP in 16-bit mode even
    though the data starts out at 8 bits. Working in a higher bit depth is
    not the same as starting out in it, but is still better than doing
    intermediate calculations with lesser precision.
    --
     
    John Sheehy, Apr 16, 2007
    #15
  16. Al Dykes

    John Sheehy Guest

    jpc wrote in
    Of course it will not deteriorate much if you don't do any levels-affecting
    post-processing.

    Adjust gamma to 0.2, then to 5.

    Increase saturation, or contrast, etc.

    You will see the image fall apart very quickly.

    --
     
    John Sheehy, Apr 16, 2007
    #16
  17. Al Dykes

    John Sheehy Guest

    The 12 bits RAW linear and 8 bits JPEG are apples and oranges. An 8-bit
    image format can have more DR than a 12 or even 16-bit format, depending on
    how the bits are used. Of course, in this particular situation, the JPEG
    can not contain any detail not present in the RAW, because the RAW is its
    origin. A camera *could* output an 8-bit format with more DR than a 12-bit
    linear, if it digitized cleanly at 16 bits, for instance.


    --
     
    John Sheehy, Apr 16, 2007
    #17
  18. Al Dykes

    ray Guest

    Quite possible that I don't know as much here as I thought I did. I have
    seen HDR (high dynamic range) photos processed from a single raw photo
    rather than several exposure bracketed jpegs. Seems to give about the same
    result.
     
    ray, Apr 16, 2007
    #18
  19. Al Dykes

    HEMI-Powered Guest

    Today, ben brugman made these interesting comments ...
    this was my point, it seems almost everyday some newbie comes
    along with what amounts to be a philosophical question about
    JPEG, usually with many misconceptions, and the elitists then
    follow with a lot of hooey that has no meaning. it is only the
    person asking the question that knows if there is or is not
    damage, not the opinion of others including them that runs lab
    testing.
     
    HEMI-Powered, Apr 16, 2007
    #19
  20. Al Dykes

    HEMI-Powered Guest

    Today, tomm42 made these interesting comments ...
    Tom, in my experience, the subject(s), lighting, exposure,
    requirement for DOF, and many other criteria affect when/if
    defects are noted in a JPEG saved even at minimum compression in
    a camera or when saving from a graphics app. But, so does the
    photographer's definition of "good", "adequate", "lousy" and
    their definition of "defects". Besides all this, there is the
    very long discussion of where and how an image will be used
    later, e.g. view on monitor, print, and if print, to what size
    (s), viewing distance from whatever, etc.

    I have found that for subjects with fine detail, such as my hobby
    of car pictures, artefacts can show up rather quickly as can
    posterization, thus I save a very conservative compression
    settings, certainly no more than 12, maybe 15 max on the standard
    1-100 scale. Then, I immediately reopen the just-saved JPEG and
    look carefully for damage that disturbs me. In about 1% of the
    cases, maybe less, I do see some damage, most often artefacts
    around the details of the car such as door handles, moldings,
    badges, grilles, etc. To remedy that, I will first try lowering
    the compression and if still evident, changing to one of 4 Chroma
    sub-sampling settings I commonly use.

    I would comment also that this is hardly a digital camera only
    issue, it is just as important and just as prevalent for those of
    us who scan continuous tone photos and half-tone printed photos
    from books, magazines, calendards and the like. Here, besides the
    general JPEG issues, there is also the mutually exclusive
    compromise of least noise with maximum detail and sharpness.

    When an OP posts a general question as here, I find that it is
    impossible to give a reasonable answer if they do not provide
    some specifics as to what they are doing, what they are trying o
    accomplish, how they intend to use the resulting images, and
    their individual tolerance for defects.

    I am certain, though, that RAW will always be superior, given
    equal conditions. However, both the 80/20 Rule and the Law of
    Diminishing Returns enters into this quickly, meaning that it can
    take more and more and more time to extract ever smaller amounts
    of improvements in what one is trying to do, thus the post-
    processing of snapshots takes a different workflow for me than do
    pictures I take at a car show or in a museum, with its own
    difficulties.
     
    HEMI-Powered, Apr 16, 2007
    #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.