Color depth

Discussion in 'Digital Photography' started by Mayayana, Aug 18, 2012.

  1. Mayayana

    Mayayana Guest

    Could someone explain to me the color depth numbers
    in use? I'm familiar with the Windows display numbers:
    It used to be that Windows could display 8-bit color, then
    16-bit color, or about 65,000 different tones. These days
    it's usually 32-bit color, which is actually 24-bit (16 million
    colors) with 1 byte for transparency data. So each of the
    RGB values has 256 possibilities. GIFs are 8-bit color. JPGs
    are 24-bit color.... So far, so good...

    Since delving into cameras I repeatedly see references
    to things like 8-bit JPGs, 16-bit color, etc. I'm assuming
    those numbers don't refer to what I understand them to
    mean, since that would imply very low-quality photographic
    images with grotesque dithering. I seem to be missing some
    information.
     
    Mayayana, Aug 18, 2012
    #1
    1. Advertising

  2. Mayayana

    David Taylor Guest

    On 18/08/2012 17:29, Mayayana wrote:
    > Could someone explain to me the color depth numbers
    > in use? I'm familiar with the Windows display numbers:
    > It used to be that Windows could display 8-bit color, then
    > 16-bit color, or about 65,000 different tones. These days
    > it's usually 32-bit color, which is actually 24-bit (16 million
    > colors) with 1 byte for transparency data. So each of the
    > RGB values has 256 possibilities. GIFs are 8-bit color. JPGs
    > are 24-bit color.... So far, so good...
    >
    > Since delving into cameras I repeatedly see references
    > to things like 8-bit JPGs, 16-bit color, etc. I'm assuming
    > those numbers don't refer to what I understand them to
    > mean, since that would imply very low-quality photographic
    > images with grotesque dithering. I seem to be missing some
    > information.


    Simply put:

    - JPEGs have 8-bit values of red, green and blue, and perhaps should
    more correctly be called 24-bit. (However, as part of the encoding
    process, the 24-bit space is reduced to a smaller number bits as each
    area of the image requires, and this together with other compression
    techniques makes the file size depend on image content).

    - 16-bit is what may be used internally by a program when operating on
    8-bit values - e.g. when multiplying to provide contrast enhancement -
    so that the precision is retained. The 16-bit result may be saved as an
    8-bit value, though, when all the processing is done.

    - a typical sensor's dynamic range can only be captured by a 12 - 14 bit
    number (4096 to 16384 values) per colour, and it is this precision which
    is saved in a "RAW" file. Converting that to an 8-bit JPEG requires
    some slight sacrifices, which may or may not be noticeable depending on
    what you are doing and on the image dynamic range.

    Perhaps there is a Web page explaining this as well..
    --
    Cheers,
    David
    Web: http://www.satsignal.eu
     
    David Taylor, Aug 18, 2012
    #2
    1. Advertising

  3. Mayayana

    Mayayana Guest

    | - 16-bit is what may be used internally by a program when operating on
    | 8-bit values - e.g. when multiplying to provide contrast enhancement -
    | so that the precision is retained. The 16-bit result may be saved as an
    | 8-bit value, though, when all the processing is done.
    |
    | - a typical sensor's dynamic range can only be captured by a 12 - 14 bit
    | number (4096 to 16384 values) per colour, and it is this precision which
    | is saved in a "RAW" file. Converting that to an 8-bit JPEG requires
    | some slight sacrifices, which may or may not be noticeable depending on
    | what you are doing and on the image dynamic range.


    So all those references to 8-bit and 16-bit must be referring
    to that processing. That would explain it. For instance, the
    UFRaw homepage has this:

    "...we circumvent the current 8-bit limitation of the Gimp, as
    UFRaw does all manipulations in 16-bits."

    It would seem that must be an extremely subtle difference,
    to manipulate an image in 16-bit, especially if the image is
    saved out of RAW before printing or other usage. (Which brings
    up another issue. Can a printer understand and print in 14-bit
    color?) But from reading discussions online I got the impression
    that an "8-bit limitation" is a big issue for professional
    photographers.
     
    Mayayana, Aug 18, 2012
    #3
  4. Mayayana

    nospam Guest

    In article <k0ok2r$7gm$>, Mayayana
    <> wrote:

    > | - 16-bit is what may be used internally by a program when operating on
    > | 8-bit values - e.g. when multiplying to provide contrast enhancement -
    > | so that the precision is retained. The 16-bit result may be saved as an
    > | 8-bit value, though, when all the processing is done.
    > |
    > | - a typical sensor's dynamic range can only be captured by a 12 - 14 bit
    > | number (4096 to 16384 values) per colour, and it is this precision which
    > | is saved in a "RAW" file. Converting that to an 8-bit JPEG requires
    > | some slight sacrifices, which may or may not be noticeable depending on
    > | what you are doing and on the image dynamic range.
    >
    > So all those references to 8-bit and 16-bit must be referring
    > to that processing. That would explain it. For instance, the
    > UFRaw homepage has this:
    >
    > "...we circumvent the current 8-bit limitation of the Gimp, as
    > UFRaw does all manipulations in 16-bits."


    but it's back to 8 once it's done. this is a big limitation of the gimp.

    photoshop and lightroom work in 16 bit or even higher.

    > It would seem that must be an extremely subtle difference,
    > to manipulate an image in 16-bit, especially if the image is
    > saved out of RAW before printing or other usage. (Which brings
    > up another issue. Can a printer understand and print in 14-bit
    > color?)


    some printers can. most don't. there also needs to be driver and os
    support.

    > But from reading discussions online I got the impression
    > that an "8-bit limitation" is a big issue for professional
    > photographers.


    it can be, but it isn't always.

    if you do a lot of adjustments, 16 bit will help avoid artifacts. if
    you do few adjustments, you probably won't notice a difference.
     
    nospam, Aug 18, 2012
    #4
  5. Mayayana

    Mayayana Guest

    Thanks. Very clear and informative. I was getting tripped
    up by the ambiguity of terms. Sometimes people are talking
    about channels and sometimes about total color depth. (e.g.,
    JPG having 8-bit color per channel, but 24-bit color as a
    format.)

    --
    --
    "Alan Browne" <> wrote in message
    news:...
    | On 2012-08-18 13:45 , Mayayana wrote:
    | > | - 16-bit is what may be used internally by a program when operating on
    | > | 8-bit values - e.g. when multiplying to provide contrast enhancement -
    | > | so that the precision is retained. The 16-bit result may be saved as
    an
    | > | 8-bit value, though, when all the processing is done.
    | > |
    | > | - a typical sensor's dynamic range can only be captured by a 12 - 14
    bit
    | > | number (4096 to 16384 values) per colour, and it is this precision
    which
    | > | is saved in a "RAW" file. Converting that to an 8-bit JPEG requires
    | > | some slight sacrifices, which may or may not be noticeable depending
    on
    | > | what you are doing and on the image dynamic range.
    | >
    | >
    | > So all those references to 8-bit and 16-bit must be referring
    | > to that processing. That would explain it. For instance, the
    | > UFRaw homepage has this:
    | >
    | > "...we circumvent the current 8-bit limitation of the Gimp, as
    | > UFRaw does all manipulations in 16-bits."
    |
    | There are several spaces to consider.
    |
    | Raw. Unadulterated (or only slightly in some cameras) from the image.
    | 16 bit per color space representing the signal, it's noise and a bit of
    | padding. (A 14 bit sensor will have about 1.5 bits worth of noise in
    | the 14 bits plus 2 bits of padding).
    |
    | Raw-conversion space. When the raw converter reads it, it may (or may
    | not) re-scale the 14 bits to 16 bits per color. It doesn't really
    | matter much. Then as you adjust the parameters of how you want to see
    | the raw file it is presented differently. The original raw data should
    | not be altered. Just the parameters affecting how it is displayed in
    | the raw viewer which are also passed to the editor.
    |
    | 16 bit storage. When saved to disk (say TIFF or DNG or similar), then
    | the parameters are applied to the data (except DNG) to modify it before
    | saving.
    |
    | Up to this point the original DR of the original capture is conserved
    | (except if user destroyed in the raw settings and saved as TIFF - say by
    | blocking up a color by excessive saturation in that channel during
    editing).
    |
    | Saving as JPG. JPG is a gamma compressed 8 bit representation of the 16
    | bit value (were it scaled up from 14 or 12), or 8 bit representation of
    | the 12 or 14 bit value. This is lossy. If you convert back there are
    | 16 bit values that can't be represented. That is to say an 8 bit number
    | can only yield 256 values out of 65536 values possible in 16 bit space
    | (and JPG doesn't even go 256 IIRC).
    |
    | eg: 16 bit color has 65536 values.
    | convert it to JPG and then red can only have up to 256 values.
    | convert it back to 16 bit, and only 256 possible numbers can be
    | made resulting in 65280 "lost" values from the original 65536.
    |
    | That's over simplified of course, but gives you a taste for the issue.
    |
    | (Considering the camera may have been 12 bits, for example, then the
    | loss is from a base of only 4096 values; 14 bits would be 16384.)
    |
    | >
    | > It would seem that must be an extremely subtle difference,
    | > to manipulate an image in 16-bit, especially if the image is
    | > saved out of RAW before printing or other usage. (Which brings
    |
    | The importance of manipulating the image at 16 bit is so that many steps
    | of manipulation result in most losses occurring in the lowest order
    | bits. The quality of the image does not reduce even after many
    | manipulations.
    |
    | > up another issue. Can a printer understand and print in 14-bit
    | > color?) But from reading discussions online I got the impression
    | > that an "8-bit limitation" is a big issue for professional
    | > photographers.
    |
    | The limitation is at capture in order to get the maximum DR and finest
    | resolution (in DR) in order to have editing wiggle room.
    |
    | Printers can accept 16 bit data but I don't know of any printer with a
    | dynamic range greater than 8 bits per color. Once you get that image
    | down on paper much of the DR is lost. The appearance of a high DR image
    | comes from careful editing prior to printing - fitting the desired look
    | of the image within the DR of paper and pigment.
    |
    | --
    | "C'mon boys, you're not laying pipe!".
    | -John Keating.
     
    Mayayana, Aug 18, 2012
    #5
  6. Mayayana

    ray Guest

    On Sat, 18 Aug 2012 11:03:31 -0700, nospam wrote:

    > In article <k0ok2r$7gm$>, Mayayana
    > <> wrote:
    >
    >> | - 16-bit is what may be used internally by a program when operating
    >> on | 8-bit values - e.g. when multiplying to provide contrast
    >> enhancement - | so that the precision is retained. The 16-bit result
    >> may be saved as an | 8-bit value, though, when all the processing is
    >> done. |
    >> | - a typical sensor's dynamic range can only be captured by a 12 - 14
    >> bit | number (4096 to 16384 values) per colour, and it is this
    >> precision which | is saved in a "RAW" file. Converting that to an
    >> 8-bit JPEG requires | some slight sacrifices, which may or may not be
    >> noticeable depending on | what you are doing and on the image dynamic
    >> range.
    >>
    >> So all those references to 8-bit and 16-bit must be referring
    >> to that processing. That would explain it. For instance, the UFRaw
    >> homepage has this:
    >>
    >> "...we circumvent the current 8-bit limitation of the Gimp, as UFRaw
    >> does all manipulations in 16-bits."

    >
    > but it's back to 8 once it's done. this is a big limitation of the gimp.


    On May 4, 2012 it was announced that the development version of GIMP is
    now capable of processing images in 16 bit and 32 bit modes - integer or
    float.


    >
    > photoshop and lightroom work in 16 bit or even higher.
    >
    >> It would seem that must be an extremely subtle difference,
    >> to manipulate an image in 16-bit, especially if the image is saved out
    >> of RAW before printing or other usage. (Which brings up another issue.
    >> Can a printer understand and print in 14-bit color?)

    >
    > some printers can. most don't. there also needs to be driver and os
    > support.
    >
    >> But from reading discussions online I got the impression that an "8-bit
    >> limitation" is a big issue for professional photographers.

    >
    > it can be, but it isn't always.
    >
    > if you do a lot of adjustments, 16 bit will help avoid artifacts. if you
    > do few adjustments, you probably won't notice a difference.
     
    ray, Aug 19, 2012
    #6
  7. Mayayana

    Me Guest

    On 19/08/2012 2:49 p.m., Eric Stevens wrote:

    >
    > It all depends. The Epson 3800 has a colour working spcae almost
    > identixcal with ProphotoRGB. Presumably the 3880 has an even larger
    > colour space. IF your JPG has an sRGB colour space, these two Epson
    > printers will comfortably handle the full colour range.


    That's from marketing speak "one hundred and x percent of sRGB". In
    reality they don't.
    Try it yourself:
    http://i47.tinypic.com/5tyzif.png
    That's the entire 16.7 million colour 8 bit RGB palette in one image
    (png compresses this down to only about 60kb - a neat trick)
    Open this in photoshop, load an Epson R3800 profile into proof setup,
    then select "gamut warning".
    Alternatively, use a program to view a rotatable 3g colour gamut
    projection, overlay sRGB and a printer profile. The effect of
    subtractive colour is that the saturated shades /between/ base ink
    colours cannot be achieved, even if there are "peaks" in the gamut
    corresponding to the primary ink colours.
     
    Me, Aug 19, 2012
    #7
  8. Mayayana

    PeterN Guest

    On 8/19/2012 5:33 AM, Me wrote:
    > On 19/08/2012 2:49 p.m., Eric Stevens wrote:
    >
    >>
    >> It all depends. The Epson 3800 has a colour working spcae almost
    >> identixcal with ProphotoRGB. Presumably the 3880 has an even larger
    >> colour space. IF your JPG has an sRGB colour space, these two Epson
    >> printers will comfortably handle the full colour range.

    >
    > That's from marketing speak "one hundred and x percent of sRGB". In
    > reality they don't.
    > Try it yourself:
    > http://i47.tinypic.com/5tyzif.png
    > That's the entire 16.7 million colour 8 bit RGB palette in one image
    > (png compresses this down to only about 60kb - a neat trick)
    > Open this in photoshop, load an Epson R3800 profile into proof setup,
    > then select "gamut warning".
    > Alternatively, use a program to view a rotatable 3g colour gamut
    > projection, overlay sRGB and a printer profile. The effect of
    > subtractive colour is that the saturated shades /between/ base ink
    > colours cannot be achieved, even if there are "peaks" in the gamut
    > corresponding to the primary ink colours.



    Is that RGB or sRGB?
    The color space for RGB is about 25% larger than sRGB.
    There is an interesting tutorial at:
    <http://www.cambridgeincolour.com/tutorials/sRGB-AdobeRGB1998.htm>

    --
    Peter
     
    PeterN, Aug 19, 2012
    #8
  9. Mayayana

    Me Guest

    On 20/08/2012 1:02 a.m., PeterN wrote:
    > On 8/19/2012 5:33 AM, Me wrote:
    >> On 19/08/2012 2:49 p.m., Eric Stevens wrote:
    >>
    >>>
    >>> It all depends. The Epson 3800 has a colour working spcae almost
    >>> identixcal with ProphotoRGB. Presumably the 3880 has an even larger
    >>> colour space. IF your JPG has an sRGB colour space, these two Epson
    >>> printers will comfortably handle the full colour range.

    >>
    >> That's from marketing speak "one hundred and x percent of sRGB". In
    >> reality they don't.
    >> Try it yourself:
    >> http://i47.tinypic.com/5tyzif.png
    >> That's the entire 16.7 million colour 8 bit RGB palette in one image
    >> (png compresses this down to only about 60kb - a neat trick)
    >> Open this in photoshop, load an Epson R3800 profile into proof setup,
    >> then select "gamut warning".
    >> Alternatively, use a program to view a rotatable 3g colour gamut
    >> projection, overlay sRGB and a printer profile. The effect of
    >> subtractive colour is that the saturated shades /between/ base ink
    >> colours cannot be achieved, even if there are "peaks" in the gamut
    >> corresponding to the primary ink colours.

    >
    >
    > Is that RGB or sRGB?
    > The color space for RGB is about 25% larger than sRGB.
    > There is an interesting tutorial at:
    > <http://www.cambridgeincolour.com/tutorials/sRGB-AdobeRGB1998.htm>
    >

    It's neither, but open it as sRGB in photoshop. I think that's default
    for images which aren't tagged anyway, but you can change this in PS
    colour settings.
    So you could open it in aRGB (or other) if you wanted.
     
    Me, Aug 19, 2012
    #9
  10. Mayayana

    Me Guest

    On 20/08/2012 3:27 a.m., Dudley Hanks wrote:
    > "Me" <> wrote in message
    > news:k0qbur$ihs$...
    >> On 19/08/2012 2:49 p.m., Eric Stevens wrote:
    >>
    >>>
    >>> It all depends. The Epson 3800 has a colour working spcae almost
    >>> identixcal with ProphotoRGB. Presumably the 3880 has an even larger
    >>> colour space. IF your JPG has an sRGB colour space, these two Epson
    >>> printers will comfortably handle the full colour range.

    >>
    >> That's from marketing speak "one hundred and x percent of sRGB". In
    >> reality they don't.
    >> Try it yourself:
    >> http://i47.tinypic.com/5tyzif.png
    >> That's the entire 16.7 million colour 8 bit RGB palette in one image (png
    >> compresses this down to only about 60kb - a neat trick)

    >
    >
    > Is that using the lossless png algorithm? Or a lossy one?
    >

    Lossless. It's a good demonstration of how well lossless compression can
    work on geometric patterns, including linear gradients.
    If you save that image in *jpg format, even at high quality, you end up
    with a file size something like 2MB, it looks okay on my monitor, but
    colour palette is reduced from 16.7 million to about 4 million. Start
    compressing it down to 500KB or less, and artifacts/banding become
    clearly visible.
    But *.png isn't very good for compressing "random" data in images.
     
    Me, Aug 19, 2012
    #10
  11. Mayayana

    Me Guest

    Me, Aug 19, 2012
    #11
  12. Mayayana

    Joe Kotroczo Guest

    On 18/08/2012 17:29, Mayayana wrote:
    > Could someone explain to me the color depth numbers
    > in use? I'm familiar with the Windows display numbers:
    > It used to be that Windows could display 8-bit color, then
    > 16-bit color, or about 65,000 different tones. These days
    > it's usually 32-bit color, which is actually 24-bit (16 million
    > colors) with 1 byte for transparency data. So each of the
    > RGB values has 256 possibilities. GIFs are 8-bit color. JPGs
    > are 24-bit color.... So far, so good...
    >
    > Since delving into cameras I repeatedly see references
    > to things like 8-bit JPGs, 16-bit color, etc. I'm assuming
    > those numbers don't refer to what I understand them to
    > mean, since that would imply very low-quality photographic
    > images with grotesque dithering. I seem to be missing some
    > information.


    Basically there's two ways of looking at colour depth: bits per pixel or
    bits per channel.

    Bits per pixel used to be the common way of looking at colour depth in
    the computer world, whereas bits per channel originates in the video
    world where they have fun stuff like chroma subsampling to reduce the
    required bit depth and the transmitted aggregate bit rate does not
    necessarily translate directly to the resulting colour depth.


    --
    Illegitimi non carborundum
     
    Joe Kotroczo, Aug 19, 2012
    #12
  13. Mayayana

    PeterN Guest

    On 8/19/2012 5:02 PM, Me wrote:
    > On 20/08/2012 1:02 a.m., PeterN wrote:
    >> On 8/19/2012 5:33 AM, Me wrote:
    >>> On 19/08/2012 2:49 p.m., Eric Stevens wrote:
    >>>
    >>>>
    >>>> It all depends. The Epson 3800 has a colour working spcae almost
    >>>> identixcal with ProphotoRGB. Presumably the 3880 has an even larger
    >>>> colour space. IF your JPG has an sRGB colour space, these two Epson
    >>>> printers will comfortably handle the full colour range.
    >>>
    >>> That's from marketing speak "one hundred and x percent of sRGB". In
    >>> reality they don't.
    >>> Try it yourself:
    >>> http://i47.tinypic.com/5tyzif.png
    >>> That's the entire 16.7 million colour 8 bit RGB palette in one image
    >>> (png compresses this down to only about 60kb - a neat trick)
    >>> Open this in photoshop, load an Epson R3800 profile into proof setup,
    >>> then select "gamut warning".
    >>> Alternatively, use a program to view a rotatable 3g colour gamut
    >>> projection, overlay sRGB and a printer profile. The effect of
    >>> subtractive colour is that the saturated shades /between/ base ink
    >>> colours cannot be achieved, even if there are "peaks" in the gamut
    >>> corresponding to the primary ink colours.

    >>
    >>
    >> Is that RGB or sRGB?
    >> The color space for RGB is about 25% larger than sRGB.
    >> There is an interesting tutorial at:
    >> <http://www.cambridgeincolour.com/tutorials/sRGB-AdobeRGB1998.htm>
    >>

    > It's neither, but open it as sRGB in photoshop. I think that's default
    > for images which aren't tagged anyway, but you can change this in PS
    > colour settings.
    > So you could open it in aRGB (or other) if you wanted.
    >


    I use ProPhoto RGB as my default color space. I also will switch to LAB
    for sharpening and color enhancement.

    --
    Peter
     
    PeterN, Aug 20, 2012
    #13
  14. Mayayana

    Me Guest

    On 20/08/2012 12:13 p.m., Eric Stevens wrote:
    > On Sun, 19 Aug 2012 09:29:03 -0400, Alan Browne
    > <> wrote:
    >
    >> On 2012-08-18 22:49 , Eric Stevens wrote:
    >>> On Sat, 18 Aug 2012 20:27:02 -0400, Alan Browne
    >>> <> wrote:
    >>>
    >>>> On 2012-08-18 20:10 , Eric Stevens wrote:
    >>>>
    >>>>> For the Epson 3800, the print quality differences between printing
    >>>>> directly from an 8-bit image versus a 16-bit image are negligible.
    >>>>
    >>>> Where 0 is the closest approximation to negligible in this case.
    >>>>
    >>>>> The Epson 3800 driver processes only 8-bit image data and hence
    >>>>> does not make use of the full 16-bit image data. A full 16-bit
    >>>>> driver could offer significantly more halftone steps than an 8-bit
    >>>>> driver, leading to smoother color gradations and tonal transitions.
    >>>>
    >>>> The printer can't even print every tone in an 8 bit per color JPG, there
    >>>> is no way it can do more with 16 bit.
    >>>
    >>> It all depends. The Epson 3800 has a colour working spcae almost
    >>> identixcal with ProphotoRGB.

    >>
    >> Working space does not mean it can actually get that onto the paper.

    >
    > I can't recall what the article said about but from what I've read of
    > Eric Chan I expect he would have measured that off the print.
    >

    I don't know where these types of claims originate
    http://www.epson.com/_alfresco/proi...Pro3880/downloads/StylusPro_3880_Brochure.pdf
    That's the Epson R3880 brochure, but while it's making claims about
    gamut exceeding what conventional wet-process prints could achieve
    (Something Epson printers have been able to do for a decade) and
    mentions the extended Gamut, I can't see anything claiming "105% of
    sRGB, 99% of prophotoRGB" or whatever. I'm pretty sure I've seen such
    claims in past Epson press releases, at first I believed it, then I
    thought about it and looked at what I could see with my own eyes.
    if Eric Chan made or repeated such a claim, then there's probably a
    mistake in interpretation of what that claim means.
    I don't have an Epson R3880, but d/l an Ilford profile for the R3880.
    http://i50.tinypic.com/ehht3.png
    That's a (reduced size) view of the full sRGB colour spectrum, with the
    Ilford Smooth Gloss profile gamut warning (in black), so with that
    paper, it doesn't cover all of sRGB.
    This is the same, but this time the image is aRGB, with obviously even
    more of the 16.7 million colours "out of gamut":
    http://i48.tinypic.com/1zn0mpw.png

    With the profile loaded, you can also toggle a soft-proof view using
    ctrl|shift|Y which will give an indication of how different the print
    will be from screen (assuming your screen can reproduce sRGB, and is
    well calibrated).
    There are probably "peaks" corresponding to the single ink colours where
    the R3880 (and other printers) exceed sRGB.
    The widest gamuts for Epson printers I've played with have been on
    Epson's premium lustre paper. Gamuts are much reduced on matte papers.



    >>
    >>> Presumably the 3880 has an even larger
    >>> colour space. IF your JPG has an sRGB colour space, these two Epson
    >>> printers will comfortably handle the full colour range.

    >>
    >> See above. And 16 bit still remains useless.

    >
    > I expect Epson knows this, but there is a gap between 8 and 16 bits.
    >
     
    Me, Aug 20, 2012
    #14
  15. Mayayana

    Me Guest

    On 20/08/2012 2:56 p.m., Eric Stevens wrote:
    > On Mon, 20 Aug 2012 13:54:41 +1200, Me <> wrote:
    >
    >> On 20/08/2012 12:13 p.m., Eric Stevens wrote:
    >>> On Sun, 19 Aug 2012 09:29:03 -0400, Alan Browne
    >>> <> wrote:
    >>>
    >>>> On 2012-08-18 22:49 , Eric Stevens wrote:
    >>>>> On Sat, 18 Aug 2012 20:27:02 -0400, Alan Browne
    >>>>> <> wrote:
    >>>>>
    >>>>>> On 2012-08-18 20:10 , Eric Stevens wrote:
    >>>>>>
    >>>>>>> For the Epson 3800, the print quality differences between printing
    >>>>>>> directly from an 8-bit image versus a 16-bit image are negligible.
    >>>>>>
    >>>>>> Where 0 is the closest approximation to negligible in this case.
    >>>>>>
    >>>>>>> The Epson 3800 driver processes only 8-bit image data and hence
    >>>>>>> does not make use of the full 16-bit image data. A full 16-bit
    >>>>>>> driver could offer significantly more halftone steps than an 8-bit
    >>>>>>> driver, leading to smoother color gradations and tonal transitions.
    >>>>>>
    >>>>>> The printer can't even print every tone in an 8 bit per color JPG, there
    >>>>>> is no way it can do more with 16 bit.
    >>>>>
    >>>>> It all depends. The Epson 3800 has a colour working spcae almost
    >>>>> identixcal with ProphotoRGB.
    >>>>
    >>>> Working space does not mean it can actually get that onto the paper.
    >>>
    >>> I can't recall what the article said about but from what I've read of
    >>> Eric Chan I expect he would have measured that off the print.
    >>>

    >> I don't know where these types of claims originate
    >> http://www.epson.com/_alfresco/proi...Pro3880/downloads/StylusPro_3880_Brochure.pdf
    >> That's the Epson R3880 brochure, but while it's making claims about
    >> gamut exceeding what conventional wet-process prints could achieve
    >> (Something Epson printers have been able to do for a decade) and
    >> mentions the extended Gamut, I can't see anything claiming "105% of
    >> sRGB, 99% of prophotoRGB" or whatever. I'm pretty sure I've seen such
    >> claims in past Epson press releases, at first I believed it, then I
    >> thought about it and looked at what I could see with my own eyes.
    >> if Eric Chan made or repeated such a claim, then there's probably a
    >> mistake in interpretation of what that claim means.

    >
    > IEric Chan is a Principal Scientist at Adobe Systems Inc., and there
    > is no way he would have made that statement on the basis what he read
    > in an Epson brochure. He has produced _many_ ICC curves for the Epson
    > 3800 and the later 3880 with a large number of paper types. That's why
    > I expect he most likely reached his conclusion on the basis of direct
    > measurement.

    He'd have to make a lot of measurements, so I hope he's got a good
    discount on Epson ink!
    You could hand pick a range of sRGB colours in the middle of the the out
    of gamut areas highlighted in the images I posted, and the printer won't
    be able to achieve them, regardless of what he says.
    sRGB is for additive colour, printers use subtractive colour, you'd need
    a very large palette of ink colours to be able to match or beat a screen.
    >
    >> I don't have an Epson R3880, but d/l an Ilford profile for the R3880.
    >> http://i50.tinypic.com/ehht3.png
    >> That's a (reduced size) view of the full sRGB colour spectrum, with the
    >> Ilford Smooth Gloss profile gamut warning (in black), so with that
    >> paper, it doesn't cover all of sRGB.
    >> This is the same, but this time the image is aRGB, with obviously even
    >> more of the 16.7 million colours "out of gamut":
    >> http://i48.tinypic.com/1zn0mpw.png

    >
    > Different papers give different results.
    >
    >> With the profile loaded, you can also toggle a soft-proof view using
    >> ctrl|shift|Y which will give an indication of how different the print
    >> will be from screen (assuming your screen can reproduce sRGB, and is
    >> well calibrated).
    >> There are probably "peaks" corresponding to the single ink colours where
    >> the R3880 (and other printers) exceed sRGB.
    >> The widest gamuts for Epson printers I've played with have been on
    >> Epson's premium lustre paper. Gamuts are much reduced on matte papers.

    >
    > Yet they often give the appearance of richer colours.


    Yes - possibly due to less reflectance.
    >>
    >>
    >>
    >>>>
    >>>>> Presumably the 3880 has an even larger
    >>>>> colour space. IF your JPG has an sRGB colour space, these two Epson
    >>>>> printers will comfortably handle the full colour range.
    >>>>
    >>>> See above. And 16 bit still remains useless.
    >>>
    >>> I expect Epson knows this, but there is a gap between 8 and 16 bits.
    >>>
     
    Me, Aug 20, 2012
    #15
  16. Mayayana

    nospam Guest

    In article <>, ray <>
    wrote:

    > >> So all those references to 8-bit and 16-bit must be referring
    > >> to that processing. That would explain it. For instance, the UFRaw
    > >> homepage has this:
    > >>
    > >> "...we circumvent the current 8-bit limitation of the Gimp, as UFRaw
    > >> does all manipulations in 16-bits."

    > >
    > > but it's back to 8 once it's done. this is a big limitation of the gimp.

    >
    > On May 4, 2012 it was announced that the development version of GIMP is
    > now capable of processing images in 16 bit and 32 bit modes - integer or
    > float.


    amazing! it's catching up to where photoshop was 20 years ago!

    photoshop gained 16 bit editing in 1992 with version 2.5 and 32 bit was
    added in 2005 with cs2. the current version is cs6.

    maybe one day the gimp will even get adjustment layers!
     
    nospam, Aug 20, 2012
    #16
  17. Mayayana

    ray Guest

    On Mon, 20 Aug 2012 15:12:13 -0700, nospam wrote:

    > In article <>, ray <>
    > wrote:
    >
    >> >> So all those references to 8-bit and 16-bit must be referring
    >> >> to that processing. That would explain it. For instance, the UFRaw
    >> >> homepage has this:
    >> >>
    >> >> "...we circumvent the current 8-bit limitation of the Gimp, as UFRaw
    >> >> does all manipulations in 16-bits."
    >> >
    >> > but it's back to 8 once it's done. this is a big limitation of the
    >> > gimp.

    >>
    >> On May 4, 2012 it was announced that the development version of GIMP is
    >> now capable of processing images in 16 bit and 32 bit modes - integer
    >> or float.

    >
    > amazing! it's catching up to where photoshop was 20 years ago!
    >
    > photoshop gained 16 bit editing in 1992 with version 2.5 and 32 bit was
    > added in 2005 with cs2. the current version is cs6.
    >
    > maybe one day the gimp will even get adjustment layers!


    In other words, you were wrong.
     
    ray, Aug 21, 2012
    #17
  18. Eric Stevens <> wrote:
    > On Mon, 20 Aug 2012 17:15:06 -0400, Alan Browne
    > <> wrote:


    >>On 2012-08-19 20:13 , Eric Stevens wrote:
    >>> On Sun, 19 Aug 2012 09:29:03 -0400, Alan Browne
    >>> <> wrote:
    >>>
    >>>> On 2012-08-18 22:49 , Eric Stevens wrote:

    >>
    >>>>> It all depends. The Epson 3800 has a colour working spcae almost
    >>>>> identixcal with ProphotoRGB.
    >>>>
    >>>> Working space does not mean it can actually get that onto the paper.
    >>>
    >>> I can't recall what the article said about but from what I've read of
    >>> Eric Chan I expect he would have measured that off the print.

    >>
    >>I doubt there is the means to measure every color possible off of a
    >>print that had ever possible color deliverable by an Epson 3800.


    > He could certainly get a close approximation to the limits of the
    > gamut.
    >>
    >>>>
    >>>>> Presumably the 3880 has an even larger
    >>>>> colour space. IF your JPG has an sRGB colour space, these two Epson
    >>>>> printers will comfortably handle the full colour range.
    >>>>
    >>>> See above. And 16 bit still remains useless.
    >>>
    >>> I expect Epson knows this, but there is a gap between 8 and 16 bits.

    >>
    >>I doubt it very much. 8 bits/color is 16M colors. Do you really think
    >>pigment ink can deliver 16M discernible colors?


    > It depends on who is doing the discerning.


    I seriously doubt whether I could get anywhere near that. I wouldn't
    be shocked or surprised if I couldn't even discern 1M different
    colours. However that doesn't stop me easily noticing a much more
    important deficit -- whether various colour spaces, no matter how
    finely or coarsely graded, can reach the awkward highly saturated dead
    spots between the chosen primary peaks. That's what gives the well
    known problems in capturing certain blues, purples, in flower colours,
    for example, especially when the image has some other colours and
    whites must must also look reaslistic along with the
    blue/purple/magenta petals.

    --
    Chris Malcolm
     
    Chris Malcolm, Aug 21, 2012
    #18
  19. Mayayana

    Mayayana Guest

    | >>
    | >>I doubt it very much. 8 bits/color is 16M colors. Do you really think
    | >>pigment ink can deliver 16M discernible colors?
    |
    | > It depends on who is doing the discerning.
    |
    | I seriously doubt whether I could get anywhere near that. I wouldn't
    | be shocked or surprised if I couldn't even discern 1M different
    | colours.

    According to this it's 10 million:

    http://en.wikipedia.org/wiki/Color_vision

    I thought I once read it was 10 million per color,
    or 10 million cubed. Either way, wouldn't that affect
    what's seen? We might not be able to consciously
    recognize all those colors, but the ability to see them
    presumably contributes to a sharpness of vision that
    would otherwise be dulled by subtle dithering.
     
    Mayayana, Aug 21, 2012
    #19
  20. Mayayana

    Martin Brown Guest

    On 21/08/2012 14:02, Mayayana wrote:
    > | >>
    > | >>I doubt it very much. 8 bits/color is 16M colors. Do you really think
    > | >>pigment ink can deliver 16M discernible colors?
    > |
    > | > It depends on who is doing the discerning.


    Ink or reflective pigment off paper is severely limited in dynamic range
    by scattered light. The highest dynamic range is preserved in Xray film
    and other transparency slides viewed in transmitted light.

    On a paper print you would struggle to see 256 shades of distinct grey
    and on most paper 100 would be more realistic. You can see about 1024
    grey levels on a monitor in a darkened room or a good Xray slide.

    Reproducing neutral grey monochrome printing accurately is actually one
    of the hardest things for modern colour printers to do.
    > |
    > | I seriously doubt whether I could get anywhere near that. I wouldn't
    > | be shocked or surprised if I couldn't even discern 1M different
    > | colours.
    >
    > According to this it's 10 million:
    >
    > http://en.wikipedia.org/wiki/Color_vision
    >
    > I thought I once read it was 10 million per color,


    No way. It is for most people with normal colour vision no more than a
    thousand distinct hues and at most a thousand levels of brightness. You
    can learn to see slightly more shades and a few people have unusual
    colour vision with a fourth type of cone. The absolute calibration sense
    of colour by the eye is shockingly bad and new colours whiter than white
    or blacker than black can be created by suitable visual illusion
    stimuli. See for example:

    http://blogs.discovermagazine.com/badastronomy/2009/06/24/the-blue-and-the-green/

    How many colours do you see? Now measure/count them in Photoshop.

    > or 10 million cubed. Either way, wouldn't that affect
    > what's seen? We might not be able to consciously
    > recognize all those colors, but the ability to see them
    > presumably contributes to a sharpness of vision that
    > would otherwise be dulled by subtle dithering.


    The eye doesn't have great resolution in colour vision. The resolution
    detail perception is in luminance only (approx the green channel).

    JPEG and TV broadcast exploit this feature by chroma subsampling the
    colour information to save bandwidth - since the eye cannot see it.

    Regards,
    Martin Brown
     
    Martin Brown, Aug 21, 2012
    #20
    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. Elijah

    epson stylus color 440 color problems

    Elijah, Sep 8, 2003, in forum: Computer Support
    Replies:
    7
    Views:
    1,632
    Blinky the Shark
    Sep 15, 2003
  2. John R. Baker

    Loss of Color Depth (Copy / Paste from Word Image)

    John R. Baker, Nov 17, 2004, in forum: Computer Support
    Replies:
    4
    Views:
    8,136
  3. Chris P in PA

    Color Depth and Repro from digicam vs film

    Chris P in PA, Dec 29, 2003, in forum: Digital Photography
    Replies:
    5
    Views:
    692
  4. AAvK

    Digital P&S and color bit depth

    AAvK, Oct 23, 2006, in forum: Digital Photography
    Replies:
    38
    Views:
    1,000
    Dave Martindale
    Oct 31, 2006
  5. Guilbert STABILO
    Replies:
    131
    Views:
    3,935
    Paul Furman
    Nov 27, 2008
Loading...

Share This Page