What Does a Printer Profile Actually Do?

Viewing 12 posts - 13 through 24 (of 38 total)
  • Author
    Topic: What Does a Printer Profile Actually Do? Read 44077 Times
  • Dave Chew
    Dave Chew
    Silver Member
    Posts: 64
    What Does a Printer Profile Actually Do?
    on: April 22, 2020 at 8:57 pm

    Soft proof and toggle the various rendering intents. See a difference on-screen? You should. Not even going about using the paper and ink simulations, you should see a difference. Check the values if you convert using differing Rendering Intents. So that begs the question, how can the display be correct/better without soft proofing compared to the print? Perhaps the display calibration isn’t all it’s cracked up to be. Perhaps the soft proofing table in the profile isn’t in sync with the output table or just not well built. Something sounds wrong in any case.

    Hi Andrew,

    Thank you for keeping up with this thread and posting. I didn’t do a good job explaining myself. I do see a difference when switching between rendering intents (when “simulate paper and ink” is not checked).

    The display is not any better without soft proofing compared to the print. the display always seems to show an accurate rendering when soft proofing. What I was trying to say is that the master file, printed through ImagePrint with no effort to make soft proofing adjustments, looks pretty darn close to what I want. If I don’t use ImagePrint and print through LR/PS, the image looks relatively flat and not so good (again, with no soft proofing effort). I can, with some moderate soft proofing effort, get the LR/PS printed image to look as good if not better than the ImagePrint version.

    My only point is the ImagePrint path requires much less dramatic soft proofing adjustments vs LR/PS. I suspect this has more to do with ink recipes employed through ImagePrint than it does profile quality. I am not trying to say I ultimately get better results through ImagePrint; simply that the magnitude of adjustments required to get the soft proof (and associated print) to look like the original file are greater through LR/PS-Epson driver vs ImagePrint.

    Dave

    How glorious a greeting the sun gives the mountains! - John Muir
    https://www.davechewphotography.com
    Find legacy Schneider-Kreuznach Apo-Digitar lens data here: https://www.davechewphotography.com/skdata/

    John Sadowsky
    John Sadowsky
    Participant
    Posts: 169
    Re: What Does a Printer Profile Actually Do?
    Reply #1 on: April 23, 2020 at 12:32 pm

    I tried to read all the comments above, but if I missed something I apologize in advance.

    As Dave says, in-gamut colors should soft-proof to the same in-gamut colors (but, for relative rendering intent only).  The problem is that almost all of our images contain out-of-gamut blacks.  Note that Dave’s list of adjustments does not include a black-point adjustment.

    We can do a black-point adjustment using the C1 level tool.  Just move the top black-point handle to the right.  (If you know the min L* for your printer/paper, put a L*a*b* readout on a pure black part of the image and move the top black-point handle to get L* = min L*.)  That adjustment should actually be a good approximation of the perceptual rendering intent, at least in the shadows.  Make that adjustment, and then compare black-point adjusted to soft-proofed versions.  I played around with it in C1 – nothing precise, just eye-balling.  Nonetheless, I could get a pretty good match between soft-proof and black-point adjusted original for a matte paper profile.

    It has been said that profiles don’t change colors – not true.  The relative rendering intent changes out-of-gamut colors by substituting the “closest” in-gamut color.  The perceptual intent brings out-of-gamut colors in-gamut and also adjusts in-gamut colors in order to preserve perceptual color difference.  (Perceptual intent is not precisely defined, and the ICC standard allows software developers to implement their own versions.)  Anyway, with the caveat that the original image probably has out-of-gamut blacks, Dave’s expectation that soft-proofing of in-gamut colors should show no before/after difference is justified only for relative rendering intent – not perceptual intent.

    Gripe:  Apparently C1 allows us to select the rendering intent only under Color preferences.  We should be able to select rendering intent in a recipe.

    More Technical Details

    One thing missing from this discussion is that color spaces have a white-point.  The PCS (Profile Connection Space = L*a*b* for print profiles) is always a D50 white point.  ProPhoto uses D50, but sRGB and Adobe use D65.  When you calibrate your monitor, you usually calibrate for a D65 white point.  (You can likely adjust that at calibration time, but D65 is recommended.)  Yet most print profiles assume D50 illumination.  Jim K mentioned the role of print illumination.  The print profile describes PCS-to-print colors for one specific illumination – usually D50.

    So how does soft proofing manage the fact that you monitor with a D65 white point, your eyes are adapted to D65, but the print profile is D50?  If the soft proofing displays exactly the print colors as is illuminated by D50 light, then (a) it will be different than the non-proofed image (even for in gamut colors) and (b) it will not appear as the print would as your eyes are adapted to D65, not D50.  On the other hand, it is most likely the case that soft proofing does a shift in white point to D65, the technical term is chromatic adaptation.  Then we’d be comparing apples-to-apples on screen, but if you hold the print up to the on-screen soft proof they will be different (and impossible to correct for as one is illuminated by ambient light and the other is a D65 backlit screen).
    <p style=”background-color: transparent; color: #0e101a; margin-top: 0pt; margin-bottom: 0pt;”>The Mac ColorSync utility allows us to examine profiles. Profiles are stored in “X/Library/ColorSync/Profile” where “X” is either the Mac HD root, or a user Profile. (You may have to look in both places.) If you just open the ColorSync tool it will give you a window showing all the profiles on your computer, and 3D displays of the gamuts in L*a*b*. Don’t do that. Go to the profile file and “open with” ColorSync. That will show you the file header and tag info. wtpt and bkpt are the white- and black-point tags. The device-to-PCS transformations are contained in the A2B tags. (B2A indicates the PCS-to-device transformation.) There are actually three transformations labeled intent-0, 1 and 2, which are in order, perceptualrelative, and saturation rendering intents. Saturation is generally not used in photography.  Printer profiles are non-linear 3D-to-3D transformation. Click on an A2B tag and the ColorSync tool will give you various slicing views into the 3D look-up table.</p>
    Click on the 3D tab and you’ll see the L*a*b* gamut.  What is interesting here is that the gamuts displayed for the A2B0 tag (perceptual) and A2B1 tag (relative) are different, with the perceptual being the larger gamut. I know Marc uses professional CM software. I wonder if he sees a similar difference in his tools.
    <p style=”background-color: transparent; color: #0e101a; margin-top: 0pt; margin-bottom: 0pt;”>You can most easily see the difference in the black-point. The black-point in the 3D chart for perceptual (A2B0) is shown as L* = 0, but for relative intent (A2B1) the black-point is L* > 0. I think this makes sense as the perceptual mapping takes PCS black (L* = 0) to device-black which has L* = min L* > 0. So the perceptual device-to-PCS (A2B) transformation, the inverse mapping, transforms device-black (= min L*) back to PCS black (L* = 0). Not so for relative intent, were all L* < min L* in PCS map to device-black. It is a many-to-one transformation. So, in the device-to-PCS (A2B1) transform takes device-black to PCS min L*, which is what the A2B1 gamut shows.</p>
    So, once again we have the question – what the heck does soft proofing do?
    <p style=”background-color: transparent; color: #0e101a; margin-top: 0pt; margin-bottom: 0pt;”>It seems straight-forward for relative intent.  In-gamut PCS color => B2A1 => A2B1 = the same in-gamut PCS color.  But out-of-gamut PCS colors are moved to their nearest in-gamut neighbor by B2A1 => A2B1.</p>
    What about perceptual rendering?  PCS black => B2A0 => device-black => A2B0 => PCS black, not the min L* of device-black, the latter being what you’d want for soft proofing.  Right?  So maybe perceptual soft proofing is B2A0 => A2B1 because A2B1 takes in-gamut device colors to their PCS equivalents without any color shift.

    JSS

    Andrew Rodney
    Andrew Rodney
    Participant
    Posts: 416
    Re: What Does a Printer Profile Actually Do?
    Reply #2 on: April 23, 2020 at 12:44 pm

    Well I for one absolutely do not calibrate for “D65”, I calibrate whatever CCT value produces a visual match (for me it happens to be CCT 5150K specifically for the backlight of my SpectraView and my viewing booth). YMMV. The only correct value to calibrate towards is the one that produces a match. For WP and Cd/m2.

    Further, no display can really produce D65 or any Standard Illuminant if you want to really be accurate. Standard Illuminants are based on actual average measurements made all over the world, with differing instruments of daylight.

    For history buffs of color:

    The series of D-illuminants was adopted by the CIE in 1971 based on 622
    measurements from the early 1960s: 249 at Rochester, NY (Kodak); 274 at
    Enfield, England (Thorn Electrical Industries); and 99 at Ottawa, Canada
    (National Research Council). Each of these labs contributed spectral
    measurements taken with different kinds of instruments measuring at
    different spectral intervals over slightly different ranges. The data
    were combined into a master set consisting of averages over 10 nm
    intervals from 330 to 700 nm from which the average and four
    characteristic vectors were calculated. The average and first two of
    these vectors account for most of the variance in the observed data and
    live on as the S0, S1, and S2 vectors used to calculate the
    D-illuminants in the CIE standard (see Wyszecki & Stiles, 2nd. Ed., page
    146). S0 is the mean, S1 provides a yellow-blue variation relating to
    cloud cover and inclusion/exclusion of direct sunlight, and S2 provides
    a pink-green variation which was thought at the time to derive from
    variations in atmospheric water vapor and haze.
    All of this was reported by Judd, MacAdam and Wyszecki, J. Opt. Soc.
    Am., Vol. 54, p. 1031 (1964) and was incorporated without change into
    the 1971 CIE standard except for the addition of the formula for
    illuminant chromaticities in terms of correlated color temperature due
    to Kelly at NBS (now NIST, Washington, D.C.).
    —–
    As for soft proofing for print, there are products that allow us to build an output profile that doesn’t default or assume D50. We can measure the viewing conditions and build that exact measurement value into the profile.
    Who said: “profiles don’t change colors?”
    Now if you assign a profile to existing data, the color doesn’t change, only the definition of the color changes but I don’t recall anyone here speaking of Assigning profiles. Nor do I recall anyone saying profiles don’t change color. What they do is define colors. The statement I made was profiles don’t correct color. Profiles as the name suggests profile,  fingerprint colors. If you convert from Adobe RGB (1998) to Epson 3880 Luster, indeed, the colors change; they have to.

    Author “Color Management for Photographers" & "Photoshop CC Color Management" (pluralsight.com)

    • This reply was modified 4 years, 2 months ago by Andrew Rodney.
    John Sadowsky
    John Sadowsky
    Participant
    Posts: 169
    Re: What Does a Printer Profile Actually Do?
    Reply #3 on: April 24, 2020 at 10:54 am

    Andrew, please, lighten up.  We are all stressed by the coronavirus lockdown.  Can we have a discussion, not a pissing match?

    I didn’t accuse anybody on this thread of saying “profiles don’t change colors” – I wrote, “it has been said …”  I apologize for not being clear.  I should not have written that sentense – it was irrelavant and out-of-place in this thread.

    Nonetheless, I believe I correctly described relative and perceptual rendering correctly.  Rendering intent can have a significant visual impact on printing and soft-proofing, so I think it was worthwhile to point out the role it plays here.  Specifically, profiles don’t change in-gamut colors under relative rendering intent, which is what Dave Chew expected.

    I also think I made a significant contribution to the thread by pointing out the role of out-of-gamut blacks.  Dave Chew’s example image on the left did not have only in-gamut colors.  It has out-of-gamut blacks.

    Of course, no display can reproduce the D65, or D50, illumination spectrum.  Displays produce RGB metamers of colors, not full spectrums.  The terms “D65 illumination” and “D65 white-point” mean different things.  The former is a spectrum.  The latter is a white-metamer, as represented by a CIEXYZ color vector.  Yes, D65 is an illumination spectrum standard.  To be precise one should say “the CIEXYX coordinates of the D50 illumination.”  Nonetheless, it is overwhelmingly it is understood that the term “white-point” is not referring to a spectrum.

    Most paper profiles assume a D50 illumination – and since I used the term illumination, I am talking about the D50 spectrum now.  A tool such as the Xrite i1 is a spectrophotometer.  When you create a paper profile, it illuminates printed swatches across the full spectrum of light.  Change the media illumination and you get a different profile, even if both illuminations have the same white-point.  Right?  Although some profiling software allows you to specify different illuminations, let’s go with D50.  With regard to the question Dave Chew asks, do you agree with the following statement?

    The purpose of soft-proofing is to reproduce the colors on the computer monitor so the viewer (who is adapted to the monitor’s white point, D65 or whatever) will be perceived as the same colors as on the print when illuminated with the D50 spectrum by an observer that is adapted to D50.

    Given that, and assuming relative rendering intent, then under those conditions I believe Dave Chew is correct to expect that in-gamut col0rs will appear the same in soft-proofing.  Jim K pointed out the importance of viewer adaptation when he said Dave C was wrong.  I’m saying that Dave is right, but only under specific viewing conditions + relative intent.  Did I get those conditions right?  If I am wrong – please explain where my reasoning failed.

    JSS

    Andrew Rodney
    Andrew Rodney
    Participant
    Posts: 416
    Re: What Does a Printer Profile Actually Do?
    Reply #4 on: April 24, 2020 at 11:30 am

    Andrew, please, lighten up. We are all stressed by the coronavirus lockdown. Can we have a discussion, not a pissing match?

    Lighten up what? I asked you who said profiles don’t change colors?” I see you can’t answer that and it isn’t a true statement (we agree on that).

    As to perceptual rendering intents; there are NO rules. The profile manufacturer can produce this intent any way they desire, just as a E6 film manufacturer can make a transparency rendering any way they desire and why they all look different (Kodachrome vs. AgfaChrome vs. Velvia). Yes, ALL Rendering intent can have a significant visual impact on printing and soft-proofing. This is why people should soft proof simply to pick the RI they visually desire.

     Specifically, profiles don’t change in-gamut colors under relative rendering intent….

    Here’s exactly what the two types of RI’s do with OOG colors (clip or compress):

    http://digitaldog.net/files/GamutClippingvsCompression.jpg

    Clipping vs compression

    The only profiles that assume D50 are those that assume D50 and are built that way. They don’t have to be built that way.

    The most important point is, not only do displays not produce D50 or D65 but that no one needs to believe they must use either to match a print next to the display and the numbers (CCT or otherwise) can range widely and the only numbers that count are those that produce a match. For me, for my viewing booth, that’s CCT 5150K and CCT 5150K is a large range of possible colors and might be differently reported by a different software package than I use.

    Author “Color Management for Photographers" & "Photoshop CC Color Management" (pluralsight.com)

    Dave Chew
    Dave Chew
    Silver Member
    Posts: 64
    Re: What Does a Printer Profile Actually Do?
    Reply #5 on: April 24, 2020 at 11:46 am

    I must say you guys have been a big help here in my understanding of how all this works. I think I have my head around how the profile really works and impacts the print / soft proof. This has always felt a bit like quantum mechanics: anyone who says they understand it obviously doesn’t understand it!

    I’ve custom set my monitor profile target is 5800. Interesting that I use the same monitor Andrew does. My eyes and viewing environment are not as well trained and/or controlled! The OOG blacks are a really good point, and probably relates to most of the adjustments I make during soft proofing with matte papers. Does anyone else fine Dehaze extremely useful in their soft proofing exercise? Perhaps I’m using it as a shortcut / crutch, but it seems to work quite well for me.

    As an aside, my Canon 4100 shows up today. Very exciting and looking forward to a fun weekend!

    Dave

    How glorious a greeting the sun gives the mountains! - John Muir
    https://www.davechewphotography.com
    Find legacy Schneider-Kreuznach Apo-Digitar lens data here: https://www.davechewphotography.com/skdata/

    John Sadowsky
    John Sadowsky
    Participant
    Posts: 169
    Re: What Does a Printer Profile Actually Do?
    Reply #6 on: April 24, 2020 at 1:18 pm

    Dave – on dehaze – what works for you works (by definition!).  Dehaze is some kind of combination of clarity and color saturation – but it is hard to know because we really don’t know the algorithm and Adobe, C1, … are not going to tell us their proprietary algorithms.  (C1 doesn’t have dehaze, but it has four versions of clarity.)  I personally would steer clear of dehaze because it is a complex adjustment with lots of unknown moving parts, and I don’t know what it is doing to colors.  I’d rather work with multiple simple tools.  So, I’d use clarity + color adjustments.

    If you are concerned with how printing changes colors, first check which rendering intent you are using – perceptual vs. relative.  Relative should preserve in-gamut colors.  If you do color adjustments and then use perceptual rendering you, you are getting adjustments on top of adjustments.  You may be doing adjustments that are just undone by the rendering, or perhaps the other way around.  For C1, perceptual is the default intent that you have to change as color preference.  I left Lr 6 months ago, but I believe there is a way to easily toggle the rendering intent within soft-proofing.

    JSS

    Andrew Rodney
    Andrew Rodney
    Participant
    Posts: 416
    Re: What Does a Printer Profile Actually Do?
    Reply #7 on: April 24, 2020 at 1:40 pm

    I’ve custom set my monitor profile target is 5800. Interesting that I use the same monitor Andrew does.

    But what viewing booth are you using? And of course, the profiles play a role.

    Author “Color Management for Photographers" & "Photoshop CC Color Management" (pluralsight.com)

    • This reply was modified 4 years, 2 months ago by Andrew Rodney.
    John Sadowsky
    John Sadowsky
    Participant
    Posts: 169
    Re: What Does a Printer Profile Actually Do?
    Reply #8 on: April 24, 2020 at 2:44 pm

    Andrew – sorry you can’t let go of the pissing match.  But here we go.

    “As to perceptual rendering intents; there are NO rules. …”  Yeah – I said exactly that in my original post – I cited the ICC standard.

    “Here’s exactly what two types of RI do with OOG colors.”  That was your response to my statement that relative intent preserves in-gamut colors.

    First, exactly what part of “in-gamut” didn’t you understand?  Second, how can it be “exactly” when we just agreed that perceptual is not precisely defined?

    Third, the jpeg showed projections to RGB spaces on the xy-plane – that’s irrelevant.  The xy-plane is a great way to compare RGB spaces, but not for printer/paper gamuts.  Those are 3D bodies, more complex than RGB spaces.  They are best viewed in L*a*b*.  xy-diagrams simply are not applicable here.  The mathematical problem of projection of a point to a body is non-trivial – clipping to an RGB cube is the exception, but not at all typical of the much harder problem of projection to a paper gamut, say, in L*a*b* space.  We don’t even have a precise definition of “nearest.”  Is it Euclidean distance in
    CIEXYX?  Is it dE in L*a*b*?  Or should we use dE2000 …

    Now I will take this opportunity to correct my statement.  The definition of relative intent is found in Section 6.2.2 of the ICC standard.  It, in fact, does change in-gamut for the chromatic adaptation, which is a transformation that maps white-point to white-point.  The equations are given in Section 6.3.2.  So, I should have said that relative intent preserves in-gamut colors between color spaces that have the same white-point.  ProPhoto, the usual working space, and the PCS both have D50 white points.

    The ICC standard doesn’t actually say anything about how out-of-gamut colors are mapped to in-gamut for relative intent.  As noted, projection to an arbitrary gamut is mathematically difficult, so manufacturers are given leeway to find their own solutions.  However they do it, the solution is embedded in the 3D lookup tables that define the PCS-to-device transformation, a 3D look-up table, stored with the profile.

    Lastly, I thought I was very clear in my last response, saying a display is calibrated to D65 means that it’s white (which is an RGB metamer) has the same CIEXYZ coordinates as the D65 spectrum.  Given that I was painfully clear, your insistence that “displays don’t reproduce D50 or D65” is just incomprehensible.  I get it!  You are talking about the spectrum.  I’m talking about the white-point metamer.  It is just so frustrating to have to keep beating around this bush on this kind of trivial technicalities!

    I’d love to have a reasonable give-and-take dialog.  That means people can make misstatements and be allowed to retract or correct – as I just did for chromatic adaptation of relative intent in this post.  You never said ‘profiles don’t change colors’ and I never said displays reproduce a D50 or D65 spectrum!  At least I apologized for my misleading statement.  If all you want to do is have a pissing battle – then I won’t be replying anymore.

    JSS

    Andrew Rodney
    Andrew Rodney
    Participant
    Posts: 416
    Re: What Does a Printer Profile Actually Do?
    Reply #9 on: April 24, 2020 at 2:55 pm

    Andrew – sorry you can’t let go of the pissing match.

    What pissing match?

    The illustration factually shows, and it isn’t irrelevant, the differences between gamut compression and gamut clipping of OOG colors with ICC profiles. What part didn’t you understand?

    The ICC doesn’t say anything about how out of gamuts  are mapped because there are no rules how to do so for Perceptual rendering intents and RelCol clips the colors to the boundaries of the other color gamut AS SHOWN. 

    deltaE is a plot, value of color distance (difference). So I don’t know why you’ve brought that subject up, it doesn’t have anything to do with illustrating the differences between gamut compression and gamut clipping.

    There is zero reason to show 3D gamut maps, of which I can easily provide WHEN a simple 2D map will and does illustrate the differences between gamut compression and gamut clipping of profiles.

    Lastly, I thought I was very clear; the numbers or targeting for WP for a display means very little; the ones that produce a visual match are all that are important and everyone’s mileage will vary. Not may, will; due to a huge number of factors.

     I’d love to have a reasonable give-and-take dialog 

    “If you want a wise answer, ask a reasonable question.” -Johann von Goethe

    Author “Color Management for Photographers" & "Photoshop CC Color Management" (pluralsight.com)

    Rand Scott Adams
    Rand Scott Adams
    Silver Member
    Posts: 287
    Re: What Does a Printer Profile Actually Do?
    Reply #10 on: April 25, 2020 at 7:18 pm

    What a GREAT discussion!   Thanks one and all for this tour de force on the subject.

    Rand

    Rand Scott Adams Rand47

    John Sadowsky
    John Sadowsky
    Participant
    Posts: 169
    Re: What Does a Printer Profile Actually Do?
    Reply #11 on: April 25, 2020 at 10:12 pm

    What a GREAT discussion!   Thanks one and all for this tour de force on the subject.

    Welt, I’m glad you appreciate it.  I thought this thread devolved into an incredibly useless CF.  For my part in that, I am ashamed.  This was an opportunity to communicate the realities of CM and soft-proofing to practicing photographers.  Instead, it devolved into a bickering battle between individuals, myself included, on technical minutia that is of no use to the practicing photographer.  And for what reason?  To be the “expert.”  Unfortunately, this happens far too often in forums on photography technology.

    JSS

Viewing 12 posts - 13 through 24 (of 38 total)
  • You must be logged in to reply to this topic.