8-bit and 16-bit images

Discussion in 'Digital Photography' started by Conrad, Jun 20, 2006.

  1. Conrad

    Conrad Guest

    Hi,

    I have a question about running Photoshop in 8-bit and 16-bit mode.
    When handling RAW images - I use the 16-bit mode for any adjustments
    (larger color gamut). When handling JPEG images from camera - that come
    in as 8-bit images - does it do any good to change these to 16-bit mode
    while adjusting them in Photoshop?

    Best,

    Conrad
    Camp Sherman, Oregon
     
    Conrad, Jun 20, 2006
    #1
    1. Advertising

  2. Conrad wrote:

    > I have a question about running Photoshop in 8-bit and 16-bit mode.
    > When handling RAW images - I use the 16-bit mode for any adjustments
    > (larger color gamut). When handling JPEG images from camera - that come
    > in as 8-bit images - does it do any good to change these to 16-bit mode
    > while adjusting them in Photoshop?


    Conrad,

    As I understand it, it will in fact do good if you're doing multiple
    adjustments to your JPEGs as there will be less precision lost to
    rounding errors.

    --
    - Michael J. Astrauskas
     
    Michael J. Astrauskas, Jun 20, 2006
    #2
    1. Advertising

  3. Conrad

    Bill Hilton Guest

    >Conrad wrote:
    >
    > I have a question about running Photoshop in 8-bit and 16-bit mode.
    > When handling RAW images - I use the 16-bit mode for any adjustments
    > (larger color gamut).


    16 bit mode does not offer a larger color gamut ...

    > When handling JPEG images from camera - that come
    > in as 8-bit images - does it do any good to change these to 16-bit mode
    > while adjusting them in Photoshop?


    No, the damage is already done when you converted to jpegs, so leave in
    8 bit mode for further edits.

    Bill
     
    Bill Hilton, Jun 20, 2006
    #3
  4. "Bill Hilton" <> wrote in message
    news:...
    SNIP
    >> When handling JPEG images from camera - that come in as
    >> 8-bit images - does it do any good to change these to 16-bit
    >> mode while adjusting them in Photoshop?

    >
    > No, the damage is already done when you converted to jpegs,
    > so leave in 8 bit mode for further edits.


    A small nuance though. If one uses several postprocessing steps, one
    also accumulates rounding errors. So if more than one simple
    correction is applied, it may help to convert to 16-bit/channel mode
    first, and finally convert to 8-b/ch again.

    Bart
     
    Bart van der Wolf, Jun 21, 2006
    #4
  5. Conrad

    Bill Hilton Guest


    >Bart van der Wolf wrote:
    > A small nuance though. If one uses several postprocessing steps, one
    > also accumulates rounding errors. So if more than one simple
    > correction is applied, it may help to convert to 16-bit/channel mode
    > first, and finally convert to 8-b/ch again.


    Yeah that's the theory ... the reality is that no one can actually
    provide an example of this where it matters ... go ahead, shoot a jpeg
    of something, make a copy of it and convert to 16 bit, run the
    identical steps on both the 8 and 16 bit versions and show us a
    meaningful difference between the two final images.

    So far as I know many have tried this experiment and all have failed to
    show an example with a noticeable difference.

    Bill
     
    Bill Hilton, Jun 22, 2006
    #5
  6. On 21 Jun 2006 16:02:57 -0700, "Bill Hilton" <>
    wrote:

    >
    >>Bart van der Wolf wrote:
    >> A small nuance though. If one uses several postprocessing steps, one
    >> also accumulates rounding errors. So if more than one simple
    >> correction is applied, it may help to convert to 16-bit/channel mode
    >> first, and finally convert to 8-b/ch again.

    >
    >Yeah that's the theory ... the reality is that no one can actually
    >provide an example of this where it matters ... go ahead, shoot a jpeg
    >of something, make a copy of it and convert to 16 bit, run the
    >identical steps on both the 8 and 16 bit versions and show us a
    >meaningful difference between the two final images.
    >
    >So far as I know many have tried this experiment and all have failed to
    >show an example with a noticeable difference.
    >
    >Bill



    It is *really* interesting and IMO quite instructive to
    experiment with bit-depths in Photoshop using the
    Image->Adjustments->Posterize tool. It takes a
    certain kind of image to look *bad* at reduced
    bit-depths.

    Bit-depth and rounding errors are most likely (in fact,
    *only*) observed in areas of minimal detail and smooth
    gradients over a narrow tonal range.

    Alas this situation does happen in landscape photos
    with skies, so the issue is real. Even in the absence
    of outright posterization, the effect is to make the
    sky (or clouds, etc.) look "grainy."


    rafe b
    www.terrapinphoto.com
     
    Raphael Bustin, Jun 22, 2006
    #6
  7. Conrad

    Bill Hilton Guest


    >Raphael Bustin wrote:
    >
    > It is *really* interesting and IMO quite instructive to
    > experiment with bit-depths in Photoshop using the
    > Image->Adjustments->Posterize tool.


    What number do you plug in to "levels" to get a certain bit depth? I
    don't think it works like you are assuming ... for example, if you
    enter '256' how many total bits of color depth do you think you are
    viewing?

    > Bit-depth and rounding errors are most likely (in fact,
    > *only*) observed in areas of minimal detail and smooth
    > gradients over a narrow tonal range.
    >
    > Alas this situation does happen in landscape photos
    > with skies, so the issue is real. Even in the absence
    > of outright posterization, the effect is to make the
    > sky (or clouds, etc.) look "grainy."


    So where is an example? Typically the rounding errors would cancel
    each other out. If they didn't, by some stroke of bad luck, I'd think
    you'd have to run many operations to get enough round-off to see even a
    couple of digits difference, which is a fraction of a bit. In other
    words if RGB 44/67/99 somehow got rounded to say 44/66/99 that would be
    a difference of one in 64 million. You can't see a difference this
    small on your monitor, on a web image or in a print. So who cares?

    The entire 8 bit vs 16 bit argument is interesting enough, but to me 8
    vs 8 converted to 16 is meaningless. For prints I work on the 12 bit
    (digital) or 14 bit (film scans) files in 16 bit mode, and most
    photographers and authors agree this workflow has occasional benefits,
    but there are a couple of guys who disagree even with this, like Dan
    Margulis. They've asked for examples of files edited in both 8 and 16
    bit mode that show noticeable differences (files that started life as
    high bit) and it's actually hard to provide examples of this, though a
    few have surfaced.

    The 8 bit tiff changed to 16 bit tiff for editing (sometimes with a bit
    of Gaussian blur added to induce noise) argument was brought up many
    years ago (say Photoshop 4 or 5 era) but every author I've ever read
    (guys like McClelland, Fraser, Blatner, Haynes, Caponigro) felt this
    was pointless, based on their testing. Basically you just don't see
    enough of a round off error to be visible with actual photos.

    Doing it with jpegs makes even less sense since you've already skinned
    off part of the image's finer points when you did the jpeg compression.

    At any rate, if you can provide an example of an 8 bit jpeg that looks
    visibly worse than the same file converted to 16 bit with the same
    edits done to it in Photoshop I'd like to see it. If you can do this
    I'll admit I'm wrong but so far as I know no one has ever supplied such
    an example.

    Bill
     
    Bill Hilton, Jun 22, 2006
    #7
  8. Conrad

    rafe b Guest

    "Bill Hilton" <> wrote in message
    news:...
    >
    >>Raphael Bustin wrote:
    >>
    >> It is *really* interesting and IMO quite instructive to
    >> experiment with bit-depths in Photoshop using the
    >> Image->Adjustments->Posterize tool.

    >
    > What number do you plug in to "levels" to get a certain bit depth? I
    > don't think it works like you are assuming ... for example, if you
    > enter '256' how many total bits of color depth do you think you are
    > viewing?


    Plug in 32 (for example) and that's 2^5 levels, or 5 bits.
    Plug in 64, it's 6 bits. I presume that's per color channel.
    A highly detailed image may look just fine with as few as
    16 or 32 levels (ie., 4 bits or 5 bits per color channel.)

    > At any rate, if you can provide an example of an 8 bit jpeg that looks
    > visibly worse than the same file converted to 16 bit with the same
    > edits done to it in Photoshop I'd like to see it. If you can do this
    > I'll admit I'm wrong but so far as I know no one has ever supplied such
    > an example.



    Probably no single operation, but a succession of operations,
    most certainly. Here's what I tried:

    1. Using Levels, adjust gamma to 1.1
    2. Using Levels, adjust gamma to 0.9
    3. Repeat steps 1 and 2 about eight or ten times.

    With this procedure, I see obvious degradation of an 8-bit
    image, but no comparable degradation of the 16-bit image.

    I am "not" claiming that this is good practice or that it
    represents real-world working conditions.

    Nor would I claim that a single conversion from 16 bits/chan
    to 8 bits/chan would "ruin" any image. That's poppycock.


    rafe b
    www.terrapinphoto.com
     
    rafe b, Jun 22, 2006
    #8
  9. Conrad

    Bill Hilton Guest

    >>>Raphael Bustin wrote:
    >>>
    > >> It is *really* interesting and IMO quite instructive to
    >>> experiment with bit-depths in Photoshop using the
    >>> Image->Adjustments->Posterize tool.


    >> "Bill Hilton" <> wrote in message
    >>
    >> What number do you plug in to "levels" to get a certain bit depth?


    >rafe b wrote:
    >
    > Plug in 32 (for example) and that's 2^5 levels, or 5 bits.
    > Plug in 64, it's 6 bits. I presume that's per color channel.


    I used to think it works that way but now I'm not sure ... plug in 255
    and if you are right you should see no change to an 8 bit/channel
    image, but I see a change. When I compare to "indexed mode" images
    (where you can plug in an exact # of colors but are limited to 256
    total or 8 bits for all three channels total) I don't see what I would
    expect either.

    If you entered "2" for levels I'd expect 8 colors if you are right (3
    bits total), but according to the Help files for Posterize (CS version)
    this is not so ... the Help file says "For example, choosing two tonal
    levels in an RGB image gives six colors: two for red, two for green,
    and two for blue."

    So either the Help files are wrong (possible) or it's not doing what
    I'd expect.

    >>bill wrote
    >> At any rate, if you can provide an example of an 8 bit jpeg that looks
    >> visibly worse than the same file converted to 16 bit with the same
    >> edits done to it in Photoshop I'd like to see it. If you can do this
    >> I'll admit I'm wrong


    >Rafe wrote
    > Probably no single operation, but a succession of operations,
    > most certainly. Here's what I tried:
    >
    > 1. Using Levels, adjust gamma to 1.1
    > 2. Using Levels, adjust gamma to 0.9
    > 3. Repeat steps 1 and 2 about eight or ten times.
    >
    > With this procedure, I see obvious degradation of an 8-bit
    > image, but no comparable degradation of the 16-bit image.


    OK, I ran a quick test with these steps and you are right, you see
    posterizing on the 8 bit image but not the same image in 16 bit mode.
    Here's the example I ran, starting with a tiff, converting to jpeg with
    8 level quality (which would make it more likely to posterize but
    that's a reasonable quality level with Photoshop's jpeg conversion),
    duplicating the 8 bit version and changing to 16 bit mode and running
    the same steps on both six times.
    http://members.aol.com/bhilton665/tests/8vs16.jpg

    So you are right ... there's a guy on the Photoshop NG (I think Mike
    Russell) who has held 16 vs 8 bit challenges in the past, with a $50
    prize for anyone providing an example showing 16 bit editing is better
    .... next time he runs the challenge I'll point it out to you so you can
    submit an example :)

    > I am "not" claiming that this is good practice or that it
    > represents real-world working conditions.


    Yes, if you have to make many edits you would be better off working on
    something other than a jpeg, but still you've made your point.

    > Nor would I claim that a single conversion from 16 bits/chan
    > to 8 bits/chan would "ruin" any image. That's poppycock.


    I'm not sure what you mean here ... I've never seen anyone claim
    converting from one mode to the other damages an image.

    Bill
     
    Bill Hilton, Jun 23, 2006
    #9
  10. Conrad

    Bill Hilton Guest

    To correct two errors in my earlier post ...

    >Bill Hilton wrote:
    > What number do you plug in to "levels" to get a certain bit depth? I
    > don't think it works like you are assuming ... for example, if you
    > enter '256' how many total bits of color depth do you think you are
    > viewing?


    Should be '255', not 256 ... 256 is illegal ...

    > In other
    > words if RGB 44/67/99 somehow got rounded to say 44/66/99 that would be
    > a difference of one in 64 million.


    Should be "one in 16 million" ... one of those days ...

    Bill
     
    Bill Hilton, Jun 23, 2006
    #10
  11. On 22 Jun 2006 17:28:03 -0700, "Bill Hilton" <>
    wrote:

    >>>>Raphael Bustin wrote:


    >> Nor would I claim that a single conversion from 16 bits/chan
    >> to 8 bits/chan would "ruin" any image. That's poppycock.

    >
    >I'm not sure what you mean here ... I've never seen anyone claim
    >converting from one mode to the other damages an image.



    I'm saying that extra bit-depth is useful to have before and
    during the major color moves. Once you stop shifting tones
    around, 8 bits per color will hold any sensible end-result.

    IOW, I think we mostly agree.



    rafe b
    www.terrapinphoto.com
     
    Raphael Bustin, Jun 23, 2006
    #11
  12. On 22 Jun 2006 17:28:03 -0700, "Bill Hilton" <>
    wrote:


    >OK, I ran a quick test with these steps and you are right, you see
    >posterizing on the 8 bit image but not the same image in 16 bit mode.
    >Here's the example I ran, starting with a tiff, converting to jpeg with
    >8 level quality (which would make it more likely to posterize but
    >that's a reasonable quality level with Photoshop's jpeg conversion),
    >duplicating the 8 bit version and changing to 16 bit mode and running
    >the same steps on both six times.
    >http://members.aol.com/bhilton665/tests/8vs16.jpg



    And notice that the effect is only apparent in the unfocused
    background. You'd never notice the difference in the
    brightly colored foreground detail.

    Roger Clark and I had this same discussion a while back
    and I was arguing your side. It became obvious after a
    few go-arounds that Roger was doing much more
    radical color moves than I typically do. If you're going
    to wail on the colors, the bit depth is useful. If you get
    the scan right in the first place and limit the extent
    (and number) of color moves, you can get by with less.


    rafe b
    www.terrapinpoto.com
     
    Raphael Bustin, Jun 23, 2006
    #12
  13. "Bill Hilton" <> writes:

    >If you entered "2" for levels I'd expect 8 colors if you are right (3
    >bits total), but according to the Help files for Posterize (CS version)
    >this is not so ... the Help file says "For example, choosing two tonal
    >levels in an RGB image gives six colors: two for red, two for green,
    >and two for blue."


    Maybe the writer wasn't counting white and black, which should also be
    available with 2 levels per colour.

    Dave
     
    Dave Martindale, Jun 23, 2006
    #13
  14. In article <>,
    says...
    >
    > At any rate, if you can provide an example of an 8 bit jpeg that looks


    I have two discussions about 8 vs 16 bits in the tips section of my
    web site:
    http://robertdfeinman.com/tips/tip3.html
    http://robertdfeinman/tips/tip27.html

    They show *differences* between the two images, whether one is
    better than the other or the differences are "noticeable" is in
    the eye of the beholder.

    I see no harm in keeping to more than 8 bits for most of the editing
    process (except for slowing down the editing) and then converting
    to 8 bit for output (print or online). For digital cameras I can
    see speed and storage space factors, but once again each person
    will have to decide for themselves. If you have an important image
    and you capture it in 8 bits or in jpeg format and sometime in the
    future you want to reprocess it to display with some miracle new
    technology, your options will be limited. If you keep all the data
    available you will have more options later.

    --
    Robert D Feinman
    Landscapes, Cityscapes and Panoramic Photographs
    http://robertdfeinman.com
    mail:
     
    Robert Feinman, Jun 23, 2006
    #14
    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. Peter Frank
    Replies:
    11
    Views:
    1,117
    Chris Cox
    Dec 13, 2004
  2. Homer J. Simpson
    Replies:
    5
    Views:
    797
    Jim Barry
    Jan 19, 2009
  3. David B.
    Replies:
    0
    Views:
    529
    David B.
    Jan 20, 2009
  4. Lanwench [MVP - Exchange]
    Replies:
    0
    Views:
    464
    Lanwench [MVP - Exchange]
    Jan 21, 2009
  5. sandi_ro
    Replies:
    2
    Views:
    575
    sandi_ro
    Jan 28, 2009
Loading...

Share This Page