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

    Sarah Guest

    First of all, thanks to all who replied to my previous question.

    My question is fairly basic (read dumb). A 16 bit RGB channel results
    in a total of 48 pixels which means you can potentially have 2.81
    trillion colours.

    8 bit gives you 16.7 million and humans can only see about 10 million,
    so is there any reason to use 16 bit per channel RGB?

    I'm probably missing something blantently obvious but any advice would
    be greatly appreciated, i.e. is it for CMYK printing purposes or am I
    completely missing the point and there is a different use for it
    altogether.

    Any help would be greatly appreciated.

    Regards,

    Sarah.
     
    Sarah, Oct 20, 2005
    #1
    1. Advertisements

  2. 8-bit/channel will give 256 colors for each channel, making
    256*256*256 colors in total. That is probably fine for a final result.
    However, if we want to do several cumulative image adjustments on low
    noise images then that can result in posterized gradients. To avoid
    that it can help to work in 16-b/ch, where accumulating rounding
    errors will have less effect on the final 8-bit/channel that's going
    to be used in final output.

    It can also help in avoiding posterization when converting between
    color space profiles (although the type of profile plays a role as
    well).

    Bart
     
    Bart van der Wolf, Oct 20, 2005
    #2
    1. Advertisements

  3. Sarah

    Bill Hilton Guest

    Sarah writes ...
    Yes, if you are heavily editing the files using 16 bit helps avoid
    posterization and other problems ... here's a very good write-up on the
    advantages and disadvantages, from Bruce Fraser ...
    http://www.creativepro.com/story/feature/7627.html ... if you find you
    don't need to make many edits then it's less useful.

    Bill
     
    Bill Hilton, Oct 20, 2005
    #3
  4. Sarah

    Mike Russell Guest

    Hi Sarah,

    Reasonable people differ on this topic, and it's worth discussing again
    every once in a while.

    In normal gamma spaces, such as are used on Macintosh or Windows, it is
    difficult or impossible to come up with an actual image that shows
    differences between being edited in 8 bit versus 16 bit. In fact, no such
    image has ever been presented to me in spite of several open invitations,
    and reward of $100. For this reason alone, I find the argument that 16 bit
    per channel data is an obvious requirement for excellent work is not
    justified.

    I have serious issues with the Bruce Fraser example that Bill refers to. It
    is based on an artificial image, and in the end it is a tautological
    argument, demonstrating only that in performing the conversion from 16 bit
    to 8 bit, Photoshop introduces 1/2 a bit of noise. I have also see people
    use a histogram to show that gaps appear in 8 bit edit that do not show in a
    16 bit edit - again this is unsatisfactory because it does not use an actual
    image.

    Bart van der Wolf has kindly presented an example, however it is, (please
    correct me if I'm wrong) a gamma 1.0 image. If you work with such images -
    and some people do - hibit images are justified. Years ago I worked at a
    place that developed hardware used to make digital movies, and 10 bits per
    channel were necessary to avoid banding in the shadows, again at 1.0 gamma.

    That said, 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.

    In the end, my feeling is the best way to decide is to use some of your own
    images - perform extreme edits on them and see if you can see banding, or
    other artifacts. If you are not experimentally inclined, and simply want to
    make every effort to assure quality in your images, you may simply decide to
    go with 16 bits, even if you cannot prove to yourself that there is an
    incremental quality benefit.
     
    Mike Russell, Oct 20, 2005
    #4
  5. Sarah

    Jim Guest

    This is especially true since no printer uses more than 8 bits per channel.
    Well, I have seen this in some real life images of mine. (No, I won't post
    any because I don't want to.) However, as near as I can tell, the effect is
    merely cosmetic, because in the end the image gets converted to 8 bits per
    channel before printing.
    Jim
     
    Jim, Oct 21, 2005
    #5
  6. Sarah

    Mike Russell Guest

    [re banding or other differences in 8 vs 16 bit images]
    Well, if you ever change your mind, I'm all ears - or eyes - as the case may
    be :)
    That's a good point. In fact, most people print at somewhat less than 7.5
    bits per channel.
     
    Mike Russell, Oct 21, 2005
    #6
  7. Sarah

    tacit Guest

    Hmm. I didn't know you'd issued such a challenge. I have such an
    image--a digital camera photograph taken without a flash under dim
    incandescent lighting. As you might imagine, the image turned out flat,
    dark, and very, very yellow. Color correction and tonal correction
    produced an acceptable image, but introduced noticeable banding in the
    shadow behind the subject.

    The shadow was not a critical part of the image, so this banding was not
    objectionable, but I suspect that in a 16-bit image it would not have
    occurred.

    I often encounter digital images that are so underexposed, for whatever
    reason, that all of the image information falls in the lower quarter of
    the histogram. Creating a normal tonal range by using the Levels command
    to stretch the histogram often introduces banding in any smooth
    gradation in the image; the same is not true of 16-bit images.
     
    tacit, Oct 21, 2005
    #7
  8. Sarah

    toby Guest

    Scepticism is good, but examples are not too hard to find; I scanned
    and adjusted a number of B&W negatives a while back that would have had
    no hope as 8-bit scans. Only the extra depth allowed me to set white
    and black points and have a decent quality result (the contrast change
    was large). Admittedly this is a rather extreme situation, not an
    ordinary decent quality image, but any of those original scans would
    fit your criterion.
    Which is actually a standard image processing trick, not a bug. Simple
    quantisation would produce considerably worse banding in some cases
    (imagine a subtle gradient). Adding one (8-bit) unit of noise allows
    the appearance of a smooth gradient between single 8-bit levels. An
    error diffusion dither would be even smoother, of course.
     
    toby, Oct 21, 2005
    #8
  9. Sarah

    Mike Russell Guest

    ....
    Hi Tacit,

    At this point, I'm not presenting this as a challenge, but I'd definitely be
    interested in seeing any image that you think might be best corrected in 16
    bits. I suspect that certain underexposed images would be good candidates
    for this.
    I'm ready to be convinced, and I think that this discussion can be
    meaningful, provided it is illustrated with images.
     
    Mike Russell, Oct 21, 2005
    #9
  10. Sarah

    Bill Hilton Guest

    Mike Russell writes ...
    I think if you read the 20 best selling Photoshop authors you'll find
    that 19 of them come out in favor of 16 bit editing for many
    situations, certainly writers like Fraser and Deke McClelland and David
    Blatner and John Paul Caponigro and Barry Haynes and Katrina Eismann.
    The 20th is Dan Margulis and even Dan finally came around and admitted
    16 bit files could be useful a few years back. So saying this is a
    controversial subject doesn't really hold up except for a few hold outs
    on the internet.
    He's using a boundary condition (gradient) to make his point on the web
    (which is not exactly set up to allow you to show subtle visual
    differences), but you can see similar problems with real life images
    that also have subtle shifts in color, especially when printed large.
    Here's an abstract image of mine that is similar to a gradient in that
    there are many subtle transitions ... I have to be very careful when
    printing it or colors block up or wash out, depending on the printer
    paper/profile used. Many of the transitions are lost just converting
    it to sRGB for web display. This is the kind of image that benefits
    from 16 bit/channel edits and careful color management ...
    http://members.aol.com/bhilton665/tulipabstract.jpg

    Bill
     
    Bill Hilton, Oct 21, 2005
    #10
  11. Sarah

    Clyde Guest

    This old topic again. <sigh> I guess I've been reading in newsgroups too
    much.

    I've also tested for my own satisfaction. I have never seen any
    difference between 8 bit and 16 bit files. Of course, I have a testing
    methodology that requires real photos viewed at real distances. I'm not
    interested in viewing anything at 400% or any magnification; that is
    completely useless.

    However, I do edit many files in 16 bit mode. I guess I'm just being
    safe. I do really twist some pictures sometimes and 16 bit may give me a
    tad more smoothness. Luckily have have a very fast computer and it isn't
    an issue. At the end they all get converted down to 8 bit.

    Clyde
     
    Clyde, Oct 21, 2005
    #11
  12. Sarah

    hoffmann Guest

    Mike,

    as far as I remember, Dan Margulis says that it is useful to
    import files with more than 8-bits-per-channel if the device
    delivers this higher resolution (12 bits, camera RAW, scanner),
    which leads to 16-bits-per-channel in PhS, because there is
    nothing between.

    After the few steps of basic corrections (levels and white
    balance) everything else can be done with 8-bits-per-channel.
    My experience.
    Gradients are not images but test patterns. Gradients should be
    dithered, anyway.

    The rasterizer (halftoning, like Floyd-Steinberg) will mix up
    everything: replace up to 16.7 millions of colors by drops of
    four inks.

    Some years ago, cute writers told us that gaps in a histogram
    would indicate a bad image quality. Totally wrong, because
    everything is mixed up by the halftoning process.

    Best regards --Gernot Hoffmann
     
    hoffmann, Oct 21, 2005
    #12
  13. Sarah

    toby Guest

    A series of corrections that modifies transfer functions (such as
    levels or curves) will almost always produce a more accurate result
    when carried out at a higher depth, because every sequential
    apply-function-and-requantise will degrade the image (less so if
    dithered, of course). The greater the precision of representation, the
    less degradation caused by each step.
    Actually it's not totally wrong; a Floyd-Steinberg (or other error
    diffusion) method if properly implemented can be regarded as a
    continuous tone approximation; discontinuities and quantisation steps
    in input will be faithfully rendered. In other words, every (R,G,B)
    combination, whether that is 8, 16 bit or any other precision, will
    produce a different halftoned balance. It is true that coarse
    halftoning can *conceal* the existence of such artefacts, but fine
    halftoning will faithfully reproduce them.
     
    toby, Oct 21, 2005
    #13
  14. Sarah

    Mike Russell Guest

    It would be more informative to see a photograph than a list of authorities.
    In any case, some of these people reference Bruce Fraser, so this is really
    a testimony to his well-deserved popularity.
    So one must be very subtle indeed, to see these differences, viz:

    Quoting from http://www.pitt.edu/~dash/type1620.html#andersen
    "One day two [weavers] came to the emperor's city [...] , claiming that they
    knew how to make the finest cloth imaginable. Not only were the colors and
    the patterns extraordinarily beautiful, but in addition, this material had
    the amazing property that it was to be invisible to anyone who was
    incompetent or stupid."
    At last, an image, and an interesting one at that. But you call it an
    "abstract image". It has been modified so extensively that it barely
    resembles a photograph at all.
    I would expect so, and I think you altered the image to achieve just such an
    "abstract" effect. Although your image is derived from a photograph, it is
    an artificial image, like Bruce Fraser's gradient. Your image is not
    relevant to the way we should edit ordinary photographs.
    Let's accept that this is "the kind of image that benefits from 16
    bit/channel ". In one second anyone, child or adult, will see that your
    example is very different from the vast majority of photographs. What can
    we conclude from this?

    Simply that you have never encountered a photograph that clearly benefits
    from editing in hibit. If you had, being one of the more reasonable and
    intelligent people in this news group, you would have made your point by
    using that photograph, instead of an "abstract image".

    If we can bear to controvert a handful of authorities, and if our images do
    not look like highly blurred and saturated abstractions, those of us who do
    most of our editing in 8 bit color, using the popular working spaces
    supported by Photoshop, can rest easily.

    Unless others express interest in this somewhat tired topic, I'll bow out
    and leave Bill to have the last word.
     
    Mike Russell, Oct 22, 2005
    #14
  15. Sarah

    Clyde Guest

    Yes, but can you see it?

    I don't argue that the mathematics would show that there are advantages
    to 16 bit in almost every phase. My argument is that I can't SEE any
    difference at 100% magnification down to normal viewing.

    If it doesn't make a visible difference, it isn't important.

    Clyde
     
    Clyde, Oct 22, 2005
    #15
  16. Sarah

    toby Guest

    This thread can be boiled down to: "In some cases, yes."
     
    toby, Oct 22, 2005
    #16
  17. Sarah

    DD Guest

    I'm sure I am not the only one following this tread
    word for word. But, enough is said for anyone to
    come to a conclusion - in fact a relieve;
    why bother about higher bit.

    Dave
     
    DD, Oct 22, 2005
    #17
  18. Sarah

    KatWoman Guest

    there will always be those photographers who get into doing extremely
    scientific methods, remember silver halide printing, the zone system and
    other methodologies for making "fine art" photos. Others of us need quick
    commercial stuff that can be printed anywhere, or have clients who want
    faster and cheaper. Some of us prefer working intuitively by 'eye" and
    always have done so. Some of us farmed out darkroom work while some of us
    slaved over our prints on archival paper with hand etching and coloring
    techniques. Some guys take their film to Walgreen's. Some of you sold and
    8x10 for $10.00 some of you charge considerably more

    What are you trying to produce as your final result? of course different
    techniques take you to different results. If you find good results in 8 bit
    why bother with 16? especially if you can't see the difference. If you are
    quite picky and find problems printing your files by all means use the 16
    bit and see if you get better results. There is room for a wide variety of
    techniques, everyone must find hi/her own niche and comfort level.

    I have seen pro photog's like Ellen Von Unwerth actually brag about her lack
    of technical expertise, her talent is in capturing a moment of feelings.
    Some of her stuff is out of focus, mostly shot on automatic 35mm with normal
    lenses. People pay her a lot of money for these poorly crafted but still
    commercially viable work. Then you have the "Ansel Adams" crowd, so
    carefully zoning by the numbers, using precise techniques, maybe making 100
    prints to get one they consider worthy. Some of us value creativity,
    freedom and ease of use over slaving techniques. Each method can produce
    good work.
    You have to work the way your muse directs.
     
    KatWoman, Oct 22, 2005
    #18
  19. Sarah

    hoffmann Guest

    Yes,

    and Nan Goldin's photos are all 'out of focus'
    but nevertheless more interesting (IMHO) than much
    of this common pizza design Photoshop stuff.

    'Pizza design' means: as many layers as possible.

    The image processing aspect (now my engineering view):
    No source image is 'true'.
    How can we expect that the reproduction quality depends
    on valid decimals (in limits) ? Especially the halftoning
    process is always an approximation how to replace color
    channel depth by a spatial distribution of ink drops.
    Much progress in the last years, but no eternal scientific
    truth.
    There are many other improvements (like super cells) or
    newer diffusion algorithms which can improve the printed
    quality for 8-bit-per-channel CMYK (which is anyway the
    standard for RIPs).

    Best regards --Gernot Hoffmann
     
    hoffmann, Oct 22, 2005
    #19
  20. Sarah

    Bill Hilton Guest

    Mike Russell writes ...
    You personally held a "challenge" a couple years back, offering a cash
    reward for anyone with an image that showed that 16 bits was better
    than 8 bits for some cases, and while most images did not exhibit this
    at least one did since you ended up paying off, right? So you should
    already have one such photo in hand. I think Dan Margulis made a
    similar request minus the reward, and got enough images showing the
    benefits of high bit edits that he changed his mind on the topic.
    No, it's not an "artificial image" with digital manipulations (which I
    think is what you mean), it's an unmanipulated photo. I shoot stuff
    like this a lot of the time, as do many other people I know and
    respect. The only "manipulation" here is in-camera, adjusting the
    plane and focus (or out-of-focus to be precise) to emphasize the shapes
    and colors at the expense of detail.
    OK ... and here's a screen dump showing the original RAW file in a RAW
    converter window (and a few others from the same memory card), with no
    digital modifications ...
    http://members.aol.com/bhilton665/screen_dump.jpg ... I think this kind
    of makes my point :)
    To me that's a good thing.

    Wife and I have an exhibit scheduled sometime next year called "Photo
    Impressionism" and all 16 images accepted for exhibition are similar to
    this (various zooms, blurs, panning, multiple exposures, etc) and none
    had any digital manipulation at all, so it's not like this is a one-of
    case. Here are a couple of links to the guy doing this type of work at
    the highest level, Freeman Patterson (I'm proud to have studied with
    him for a week in New Brunswick, he's the most creative photographer
    I've ever met) ... his book "Photo Impressionism and the Subjective
    Image" is a good source for these techniques and he is strictly a film
    guy with no digital work at all ...
    http://www.freemanpatterson.com/prints/impressionism1.htm
    http://www.freemanpatterson.com/prints/impressionism2.htm

    Bill
     
    Bill Hilton, Oct 23, 2005
    #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.