Can't figure out 16 bit per channel RGB Channel in PhotoShop - NEWBIE

Discussion in 'Photoshop Tutorials' started by Sarah, Oct 20, 2005.

  1. Sarah

    Mike Russell Guest

    From: "Bill Hilton" <>

    [re 16 bit mode]
    The first part is true. Three years ago, I did issue a 16 bit challenge.
    Designing such a challenge was more difficult than I imagined. Although I
    gave the final prize to Michael Schaffer, I did not end up with the
    hoped-for image that does better when edited in 16 bits per channel than 8

    I'll refrain from responding in detail to the rest of your post. I was
    mistaken in thinking that your images were manipulated in an editor. This
    does not change the gist of my argument, and raises the question of where
    the benefit of 16 bits is, if you did not edit the image in question.

    Your images are attractive works of art, and I expect that they will be well
    received. Best of luck to both of you with the show!
    Mike Russell, Oct 23, 2005
    1. Advertisements

  2. The very fact that this discussion (argument??) is taking place
    suggests the (as yet) uncertain value of 16 bit over 8 bit. Given the
    "givens" , at the very least, working in 16 bit simply guarantee
    maximum (potential) quality. Ken
    Ken Palmateer, Oct 23, 2005
    1. Advertisements

  3. Sarah

    Bill Hilton Guest

    I'll try to show an image that does illustrate the advantages of 16 vs
    8 bit edits below but first ...
    No problem.
    I showed that particular image because you were challenging Fraser's
    use of gradients to make his point ... the image I showed has natural
    gradients similar to what Bruce used, albeit over a much narrower color
    range. My point was that such gradients do occur in photos at times.
    Since my image was exposed properly it requires little editing, but
    because of the saturated reds it looks much better converted from the
    RAW into a wider gamut space like Ektaspace and then converted to a
    tighter gamut profile with perceptual rendering, so for that reason I
    would want to keep it in 16 bit mode. Had it been over or underexposed
    you would probably see what Fraser was showing with his example.

    Anyway, here's another image you might find of interest ... I spent the
    first 18 days of Oct in New England hoping to photograph fall foliage
    but it was a terrible year, the first week had record heat (mid 80's
    instead of 50's where I was) which killed the colors, then we had
    record rains and flooding, which didn't help the photography.

    I got this shot of a loon in a heavy fog one morning, I wanted a dreamy
    effect with just a hint of a shape to catch a mood ... ... this is unedited
    from the RAW, shot at + 1 2/3's off meter reading, and is still a bit
    dark. You can see from the histogram that only about 10% of the tonal
    range is used, with the bulk of the data between 214 and 235 or so.
    The light was so low contrast that an f/2.8 lens would AF but a f/4
    lens would not.

    So for this thread I converted twice, once at 8 bit/channel and once at
    16 bits, and then set the black point on the loon's neck and the white
    point on a bit of foam floating near the bird, basically spreading 30
    data points across 245 points (I had the white point set for 245).

    Here is the 8 bit result ...
    Here is the 16 bit result ...

    Two concessions before you say anything ... 1) yes, this is an extreme
    tonal change and 2) both look like crap at 100% ... no argument from me
    on those points, but to me the 16 bit version looks less crappy with
    less noise. And from the histogram you'd infer that further edits
    (perish the thought) would cause less damage with the 16 bit version.

    Now while you (or at least I) wouldn't make such a radical tonal shift
    with this image (I like it the way I shot it) I did get asked to darken
    the bird when I posted this image on a web site ... here's the image I
    posted, created from an 8 bit conversion btw ... (click 'back' to see
    others from the trip that are not so foggy :) ... the first
    person to view this told me they could hardly see the bird and I needed
    to adjust it, so if I were to do this I'd make a less drastic move of
    the black point than in the example, but still a fairly major movement.
    At some point you start to see the difference between making an edit
    like this in 8 bit vs making it in 16 bit. Exactly how far you have to
    move the black point before it becomes noticeable is something I'm not
    going to check, but at some point ... if I'm lucky enough to sell some
    large prints of this image I'll certainly do the prep work on a 16 bit

    Another poster in this thread wrote that some people are willing to go
    to more trouble for small quality improvements and I'd plead guilty to
    that ... if I could get a 5% improvement for a 100% increase in effort
    I'd call it a bargain. I know others don't feel that way but I do.

    One last image, this time without the before/after ... my wife shot
    this hummingbird earlier this summer and the image was about 1.5 stops
    underexposed because she had been shooting in a different direction
    with minus exposure compensation set (needed + for this light) and
    because the fill flash didn't re-cycle quickly enough for the flash to
    be fully powered (something about shooting at 8 frames/sec when the
    flash needs up to 1.5 sec to recharge) ... ... the tonal range
    wasn't as limited as the foggy loon shot but it was still compressed,
    so when someone wanted an 11x14" print from a cropped area it looked
    smoother (in the print) going back to a 16 bit file than it did using
    the 8 bit version I had originally converted on the laptop for the

    The common thread in both of these examples (one high key, the other
    underexposed) is that you have a limited tonal range that gets expanded
    with heavy edits. To me that's when 16 bits is better than 8 bits. To
    each his own.

    Bill Hilton, Oct 23, 2005
  4. Sarah

    Mike Russell Guest

    Photoshop CS now offers a 32 bit per channel format. Not all editing
    functions are available, as was the case initially with 16 bit per channel.
    Would you suggest changing over to 32 bit when possible, using the same
    Mike Russell, Oct 23, 2005
  5. Well, somewhere along the way, the point of diminishing returns will
    have to dictate where to draw the line. Can we see the diff between 8
    and 16? Or is this a case of were we've got to climb the mountain
    just because it's there? Ken
    Ken Palmateer, Oct 23, 2005
  6. SNIP
    HDRI is not about more precise tonality, its 32-bits/channel are
    needed to encode huge dynamic range. The numbers needed dont fit in a
    16-bit integer number. The resulting tonemapping challenge is how to
    fit that dynamic range in the limitations of 8-b/ch output modalities.

    Bart van der Wolf, Oct 23, 2005
  7. Sarah

    Mike Russell Guest

    [re Fraser's use of gradients to illustrate the hibit advantage]
    I'll stick with my contention that Fraser's article is flawed. In the
    interest of not repeating myself, let's accept, for now, Fraser's use of
    artificial images, as part of an experiment to illustrate, in an abstract
    way, the concept of 8 versus 16 bit manipulations.

    Since one of the most basic requirements for an experiment is
    reproducability, I think it is reasonable to ask that we be able to
    recreate the result ourselves.

    The Fraser article does not provide the original images, or any description,
    that I could find, on how to recreate those images. So I created an image
    that resembles the one Fraser used in his article, using a horizontal
    spectrum gradient combined with a vertical black/white gradient in
    luminosity mode.

    Here's the surprise: I applied the same levels manipulation described in the
    article. The result: there is no visible difference in the appearance of
    the 16 and 8 bit results.

    This took me all of 10 minutes to do, and I invite anyone reading this
    article, to do so themselves. You will need a version of Photoshop that
    supports 16 bit gradients.
    Again, an interesting image.
    Both links point to what appears to be the 16 bit image. Earlier this
    afternoon both of them pointed to the same 8 bit image. While you are
    fixing this, I'd like to know if you did your editiing in a wide gamut space
    such as Ektspace?
    Re the histogram, I reject it utterly as a sole measure of image quality.
    If there are problems it should be visible in the image as well as in the
    histogram. Furthermore, Photoshop injects fine noise when converting from
    16 bit to 8 bit, and this added noise will "fatten" the histogram, and make
    subtle transitions smoother.
    This is your rightful decision about your own technique. Given your obvious
    excellent results, I respect your decision, even though it appears, so far,
    not to be based on the appearance of the image, but on your own estimation
    of the underlying quality. I do not trivialize this. I believe it is the
    right of every artist and craftsman to make many such decisions in the
    course of their work.

    But, if you are going to recommend that others use hibit images for their
    editing, I think you should justify it by showing an actual image where you
    believe it makes a difference. Perhaps the loon image does this - I haven't
    seen it yet.
    Again, a beautiful image, but without the ability to compare the 8 bit and
    16 bit manipulations it does not advance the contention that 16 bit edits
    give superior results.
    This may be the case, but without comparison images, we are reverting to
    talk. You did provide comparison images, and for the life of me, I cannot
    tell the difference between your 8 bit and 16 bit versions. Once again, I
    would like to compliment you on the quality of your and your wife's images.
    The hummingbird image in particular is incredible!
    We end in agreement. I quote from my first post to this thread:

    "... many people, including some excellent professional photographers,
    feel that even an imperceptible difference in quality may eventually be
    significant to their images - if not now, then perhaps in the future. As a
    tool provider, this is enough reason for me to fully support both 8 and 16
    bit manipulations in Curvemeister."

    I would add to this that the work of many of these individuals who choose to
    work in 16 bit, as is the case with yours, Bill, is of excellent quality.
    As the evidence stands, I feel justified in asserting that there is no
    advantage to editing in 16 bits versus 8 bits.

    Once again, I'm speaking of color images in the color spaces commonly used
    in Photoshop. Monochrome images, linear gamma, and some of the more exotic
    color spaces, such as ProPhoto RGB, are a different issue.
    Mike Russell, Oct 24, 2005
  8. Sarah

    Bill Hilton Guest

    Here is the 8 bit result ...
    I only posted them once and clicking the links shows two different
    images to me, so I don't know what you are seeing or why you are having
    problems with your browser ...
    This one was edited in AdobeRGB and converted to sRGB before 'save for
    the web', which is my normal flow for most images... I typically use
    Ektaspace when the colors are highly saturated and I'm going to print
    it (instead of going to the web), like the abstract tulip, but with a
    low contrast, low saturation image like this one it is pointless to use
    a wider gamut space. I know many people would now use ProPhoto RGB
    instead of Ektaspace since it's an even wider gamut space but I'm used
    to Ektaspace from my Velvia film scans (the gamut was created by Joe
    Holmes to match E-6 transparency film) so I still rely on it for
    certain images.

    As for the rest of your comments I don't have anything else to add that
    hasn't already been said several times :)

    Bill Hilton, Oct 24, 2005
  9. Sarah

    Mike Russell Guest

    The 8 and 16 bit versions of the images are now showing up correctly.

    The 8 bit version also has a strong yellow cast compared to the 16 bit
    version that I cannot explain in terms of an 8 vs 16 bit difference. For
    the time being I have to conclude that there is something else going on in
    addition to any 8 versus 16 bit . Possibly the raw converter uses a
    different algorithm to convert 8 and 16 bit images.

    If you can provide me with a cropped version of the 16 bit image, I would
    like to try my own experiments. Otherwise, I'll simply say this is an
    intriguing image that may indicate a difference in editing capability
    between 8 and 16 bits.
    No problems with those color spaces.
    Perhaps someone will be intrigued enough duplicate my efforts to try to
    recreate Bruce Fraser's results in his article:

    Take care. You've made some valuable contributions to the discussion.
    Mike Russell, Oct 24, 2005
  10. Sarah

    JFV4492 Guest

    Working in 16 bits will also guarantee:

    - increasing the file size by several folds. After adding a few layers,
    a high resolution image file can quickly be gigabytes in size.

    - PS operations will dramatically slow down as the file size increases.

    - Large files occupy lots of hd space.

    As you questioned in a subsequent post, not all images or operations can
    benefit from working with 16 bit. Even if the benefits can be expressed
    in math or histograms, they may not be observable to the average (or
    even expert) viewers. Images intended for web, or sent to a printer are
    all converted to 8 bit. Those who choose to work with 16 bits typically
    are working with high resolution files, and will pay the price mentioned
    JFV4492, Oct 24, 2005
  11. Sarah

    Bill Hilton Guest

    Mike Russell writes ...
    I cropped out 600 x 300 pixels from each version (8 and 16) and can
    email both to you if you want to play around. About 1 MB for the 16
    bit one, half for the 8 bit one so not a problem to email them. This
    has the loon (where I set the black point (0/0/0), on the neck) and the
    bit of foam I used to set the white point (to 245/245/245) so you can
    duplicate what I did.

    I don't think I saved your email address from the last time we
    exchanged files so if you could email me I'll send these in a reply ...
    To me this just looks like added noise in the 8 bit version ...
    I used Capture One LE, the files look pretty much identical before
    editing, though I didn't do any numerical analysis on them.

    Bill Hilton, Oct 24, 2005
  12. Sarah

    Bill Hilton Guest

    JFV4 writes ...
    The files double in size, which is not "several folds" ...
    Speaking from experience, this was a real problem with medium format
    sized film scans but is not a problem with typical digital camera sized
    files. My 6x4.5 cm scans were 340 and 170 MB (16 and 8 bit), which was
    doable, but my 6x7 cm film 16 bit files were 550 MB and these did
    indeed choke my computer, even with 2 GB of RAM and a high allocation
    to Photoshop.

    In comparison, now I'm working with a lot of 8 and 11 Mpixel files,
    which are 48 and 66 MB in high bit mode (smaller than even 35 mm film
    scans). These hardly slow the computer down at all since I'm always
    working out of RAM.
    With digital the RAW files are archived and they are a fraction the
    size of high bit tiffs or psd files, so this is not that big a problem.
    Also, internal hard drives are costing less than 50 cents/GB and
    externals are often on sale for a buck a gig so disk space is not the
    problem, getting excellent images worthy of keeping on the disk is the
    problem :) I have two 200 GB internal drives and five external HDs for
    multiple backups and don't really see a problem with keeping a small
    subset of my files in 16 bit mode that are going to be converted and

    Basically what I'm saying is it's much easier to store and work on
    digital files in high bit mode than it is to work on high rez film
    scans due to the relative size of the files.

    Bill Hilton, Oct 24, 2005
  13. Sarah

    toby Guest

    Compression is typically less effective on 16 bit images (such as the
    RLE Photoshop uses for PSD). This could account for a >2-fold increase
    relative to a (compressed) 8 bit file.

    toby, Oct 24, 2005
  14. SNIP
    Which indicates there is a lot of "detail" at the Least Significant
    Bit(s), even at 16-b/ch ...
    It may be noise, it may be grain, it may be (e.g. color space
    conversion) dithering, but it is part of the image. Accumulating that
    "detail" in the LSBs of an 8-b/ch image only makes sense after all
    corrections have been done, not before.

    The remaining question is; will it show in print/on screen (whichever
    the purpose)?
    That is hard to predict, because there are many factors that can hide
    it, but one thing is clear. If it shows, it is too late to remedy,
    unless one starts all over again.

    Bart van der Wolf, Oct 24, 2005
  15. Sarah

    johnsailor Guest

    Please excuse me if the info below has already been cited.
    This article by Mike Chaney (Quimage author) very clearly addresses
    this question, and strongly suggest that 16 bits can be very useful,
    especially, as cited below, in instances of underexposure. He also
    stongly suggests using "raw capture mode" if your camera supports it.

    Here is the url :
    johnsailor, Oct 25, 2005
  16. Sarah

    Mike Russell Guest

    Good point. I've experimented with this myself, and raw files can be very
    useful for rescuing underexposed images, and for simulating a higher ISO

    Whether editing in 16 bits is superior is a separate, and IMHO still open,
    Mike Russell, Oct 25, 2005
  17. Sarah

    toby Guest

    Isn't the answer simply "it depends"?
    toby, Oct 25, 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.