11 Replies Latest reply on Jun 18, 2009 10:56 PM by Guillaume Muller

    ICC Profiles Clipping Gamut

    PhotoButler

      I am requesting information on a general topic: When I create an ICC profile whether through a profiling engine built into a rip or a stand alone application such as Monaco or Profile Maker Pro, I find that the gamut of the output device is clipped.

       

      A printer, media and inkset combination which can for example produce a vibrant red, is restricted by the ICC profile and the result is mutted.

       

      Please do not respond with any theory about device dependent and independent color.

       

      The source profiles and rendering intents are noticably affecting the output, but I am struggling to achieve a good gamut of what the printer is actually capable of producing.

       

      On many occasions customers have elected to bypass ICC profiling particualrly on the vector components of images for this reason.

       

      If anyone has any practical advice on how to maximize color gamut within an ICC workflow, I would appreciate hearing it.

       

      Thank you,

       

      The PhotoButler.

        • 1. Re: ICC Profiles Clipping Gamut
          Lou Dina Level 3

          PB,

           

          First....how did you establish the fact that the gamut was "clipped"? Are you printing RGB files, CMTK files, etc?  Are you soft proofing before printing, or just comparing a raw rgb file to printed output?  Are you printing to low gamut matte fine art papers or high gamut photo papers?  You probably know this, but gamut is a function of printer/paper/inkset combined.

           

          I usually test for the best driver settings (if using a driver), then print my target using those settings.  If using a RIP, set ink limits, linearize, then print the target.  I let the target dry at least overnight, then take readings and generate a series of profiles.  If you want to test your profile, print an RGB target, such as the Greatag 283 patch RGB taget, or a CMTK target, using relative colorimetric, and see how the super bright, OOG colors reproduce.  If the profile was properly built, you should get pretty much the same results on all your bright colors, ie, reds, greens, blues, yellows, as in your original target. An ICC profile is only as good as the target you print.

           

          Whenever you try to cram a larger color space into a smaller one (most print spaces are generally smaller, though there is overlap) you will definitely see a reduced  gamut.  RelCol rendering will keep most of the in gamut colors the same, but will remap the OOG colors to the closest "printable" color (based on the algorithm used to generate the profile).  Perceptual will scale all colors to fit into the new color space, sacrificing accuracy for detail and relationships.

           

          With Monaco Profiler, I usually set perceptual options for a +25 saturation, and leave the gray axis and contrast set to zero.  To me, these give the best compromise between detail and saturation when using perceptual intent.

           

          ProfileMaker provides much less control, at least with RGB profiles.  Chroma Plus seems to give more accurate colors, but at the expense of saturation and luminosity shift.  Colorful usually retains more saturation and has less luminosity shift, but sometimes does funky things to certain colors and details.  One size does NOT fit all.

           

          For CMYK, I prefer Monaco Profiler and use 75% GCR on the slider, 25% black start (ie, 75% black luminance), 100K and 320 TIL for sheetfed work, Intelligent Black is checked, and set the Perceptual saturation to +25.

           

          If you are printing on Matte fine art papers with an inkjet, expect to be disappointed when you compare your monitor to your print.  Same with uncoated stock on press.  Most matte papers have a limited Dmax and color gamut, and you will have some serious remapping.  Glossy, semigloss and luster photo papers have a wider gamut and much better Dmax, and so you will see less shift.  When put behind glass, these differences diminish, and the viewing light also makes a big difference.

           

          Remapping a large source space to a smaller output space is a difficult challenge.  What works great for one image often doesn't work very well for other images.  I prefer to adjust my images while still in RGB to bring them into the gamut of the output space BEFORE conversion.  Once in my output space, I usually tweak the file (if I consider it worth the effort) before printing. This way, I can maximize the saturation and contrast while retaining detail.

           

          I usually print vector images with Relative Colorimetric intent and BPC.  If when designing your vector art, you select colors that are outside of your destination gamut, then you will get compression. I just select colors that fit into my destination space.

           

          Hope something here strikes a chord.

           

          Lou

          • 2. Re: ICC Profiles Clipping Gamut
            PhotoButler Level 1

            Good morning Lou,

             

            All of your points make sense and I am fully congnicant of these issues.

             

            I was remiss in not being specific enough in my initial description. My job

            is to setup workflow for large format digital printers. I work with Onyx,

            Wasatch , Caldera, Ergosoft and other rips.

             

            Not withstanding the fact that we know the way to optimize color is to work

            in RGB until the point of conversion to an output device, all our clients

            are in a CMYK workflow.

             

            My issue with profiling is as follows: When I am working with Ink

            restrictions, linearization and total ink limiting, we are achieving far

            more saturation in and vibrancy in CMYand K. Once an ICC profile is

            introduced I can see even without printing anything, just looking at the

            input and output numbers that the gamut is clipped.

             

            When I say gamut clipped I am refering to the fact that an input of 100M

            becomes for example 70m and 3C. or variation thereof.

             

            For this reason many of my customers are bypassing ICC altogether and

            tolerating the very grainy appearance of an vector file for the gains made

            in gamut.

             

            I am able to increase the gamut of the ICC workflow with the use of

            alternate source profiles and rendering intents. But I can not achieve the

            vibrancy the ink paper driver and printer combination will produce in a raw

            or machine state.

             

            Thanks again,

             

            Steve Shaw

            • 3. Re: ICC Profiles Clipping Gamut
              Lou Dina Level 3

              Steve,

               

              Thanks for the additional info.  Now I know you are experienced with color management, RIPs, etc.  It would be helpful to give an example of a CMYK to CMYK conversion, with source and destination profile, rendering intents, BPC on or off, program you are using, etc.

               

              So far, the only way I have found to get such a lousy result (ie, 100M  becoming 70M/3C or therabouts), was as follows.  I created a new CMYK file in Photoshop, then created gradients for each primary and subtractive color, blending to white.  I posterized each color and created an 11 step step wedge from full saturation to pure white... 100C, 100M, 100Y, 100K, Red (100M+100Y), Green (100C+100Y), Blue (100C+100M), black build (100c/100M/100Y), and Rich black (100 of each ink). This gave me a CMYK reference file with full range of tones and saturations.

               

              I assigned ISO Newsprint 26v4 to the original CMYK file, (a very small container, indeed), and then converted to Euroscale Coated, Relative Colorimetric, with BPC OFF (not recommended).  The colors totally wash out, as expected.  In this example, 100M in the original becomes 69M/2Y.  The other colors take a similar hit and look terrible.  As I am sure you know, 100M on newsprint is a very weak, unsaturated color, so to give you that same drab color after conversion (in Lab terms), the wider gamut Euroscale coated profile has to use less Magenta. Turning BPC on makes a significant improvement, but you still end up with a weak result, since the source space was so limited.

               

              You are clearly experienced and knowledgeable, so I hesitate to suggest such obvious things.  What are the color settings in the various programs being used (I am assuming Adobe InDesign, Illustrator, Photoshop, etc).  It would help if the color settings are coordinated across programs and set to a suitable CMYK space.  And it is worth checking the color policies.  With such dismal results, it sounded as if your customer might be designing for newsprint, or either purposely or inadvertently assigning a low gamut profile as the source (which is why I tried the above experiment).

               

              I am not necessarily recommending the following as a workflow, but it might uncover some issues.  Take a customer CMYK file that gives you this terrible result.  See what profile the file is tagged with (or is it untagged?).  If untagged, assign a fairly wide gamut CMYK profile, such as for a sheetfed press on coated stock.  Convert the CMYK file to a reasonably wide gamut RGB space, such as Adobe RGB, then convert to your digital press profile.  Use Relative Colorimetric and BPC for all conversions.  Do you still get terribly washed out colors? I wouldn't think so.

               

              I'm sure you know that if the CMYK file supplied by your customer is untagged, then opened in Illustrator or Photoshop, those programs have to ASSUME a color space and will assigned one, based on the default color settings and policies in the Color Settings menu for that program.  Maybe there are some unwanted conversions happening unawares due to incorrect policy settings?

               

              I am grasping at straws.  A more detailed explanation of what you are doing, step by step, may reveal more.  I have to work hard to get such a bad result, and I can understand why you're looking for another approach. I have sent a lot of jobs to press and have had excellent results using ICC workflows.  I have even converted CMYK file to a differency CMYK profile (device links are usually better), and even from CMYK to RGB for editing, and back to CMYK.  I try to avoid that, but sometimes it is the only good solution.

               

              BTW, have you inspected your digital press profile in a program, such as ColorThink, or another gamut viewer?

               

              Regards,

               

              Lou Dina

              • 4. Re: ICC Profiles Clipping Gamut
                bbeens

                Steve-

                 

                One test I would suggest is print the file you are having issues with but without the ICC file in your rip.  Leave your individual ink reductions and total ink limit set as per your profiling.  If this produces the same unhappy results you can forget about the profile being the cause.  Step back and take another look at what you set the ink reductions and total ink limit in the rip.  If the results are better, in saturation only (very important: color accuracy will not be correct with this test) then look at the UCR/GCR settings in your profiling software.  Or reprint your patches and scan then regenerate your .icc/.icm file.

                 

                The biggest effect on an imput value of 100% M turning into output of 70%m 3%c will be your ink reductions / ink limit.  I would guess your M is reduced to 70% and the .icc/.icm file is adding the 3%c to account for media cast.  Complete guess there, take it for what it is worth.

                 

                Bryan

                • 5. Re: ICC Profiles Clipping Gamut
                  p taz Level 3

                  I don't think anyone has mentioned a couple of the workflow considerations here... pardon me if I overlap with anyone, I came in a bit late...

                   

                  (non expert personal opinion follows... If you have profiled a device and used the resulting profile as a gamut for design, then the designer can use the entire gamut by colouring within that gamut and producing an output file that specifies that device as the destination, then no conversion will take place and you will get what you asked for.

                   

                  If you are working to this spec and not getting what you asked for, then an unnecessary conversion is taking place.

                   

                  If your designer is designing to ANY other condition and converting then you WILL get a conversion and your output, by definition, will not use the entire target device gamut.

                   

                  Every conversion has to use less than the entire gamut of the target profile.

                   

                  Reason:

                   

                  1. The target profile must be bigger than the source or it would clip the source.  (Common for inkjets etc used for proofing for various presses etc, it must be bigger to accept more input sources)

                   

                  2. If the target and source are the same, no conversion is needed, hence no clipping.

                   

                  3. If the entire source gamut fits within the target then the source still needs to be converted to the target in order to be printed.  (This is what throws people who believe that 'turn off colour management' buttons actually do that.)

                   

                  You should be able to spec any colour within your device gamut and print it... if you cannot, then you have a workflow problem or a software problem.

                   

                  You are stating that your new icc profile is restricting the device, so anything I have said about the target is really referring to the device icc, whether your new icc profile uses the entire device capabilities is another question!  But the point stands, if you design to the gamut, then those colours should print without conversion.

                  • 6. Re: ICC Profiles Clipping Gamut
                    PhotoButler Level 1

                    Bryan,

                     

                    I understand exactly what you are saying. In my particular case I have not

                    restricted the M or C channels at all for the very reason you describe. If I

                    print without the ICC profile I can achieve 100% of Magenta. Visually it is

                    not satisfactory but this is beside the point and is due to the particular

                    ink set and oven/ transfer function of this particular fabric printer. But

                    in terms of the rip the color is read as 100 M.

                     

                    Once I introduce the ICC profile the transform from source to output is

                    remapping the pure hue. I a good understanding of rendering intents, device

                    dependence and independence and the fact that this event can be explained as

                    the Magenta is being remapped to absolute values based on LAB rather than

                    the spectral values of that particular ink.

                     

                    When I change the source profile from USWebCoatedswopV2 to a Gracol profile

                    or others which I do not have before me at the moment, the values of the

                    Magenta are significantly superior. My conclusion here is that the source is

                    clipping the destination. The other issue is that Perceptual rendering,

                    while giving me generally pleasing color in this particular rip (Onyx) is

                    detrimental to the gamut. If I use relative colorimetric while introducing

                    some issues in raster images, it is pushing the gamut and is consequently

                    more desirable for the vector elements.

                     

                    I posted the question because I have observed throughout my years in color

                    managent that we experience many compromises with ICC workflow. I would like

                    our community to acknowledge some of the limitations. This issue that I have

                    described has been common to my experience with Ergosoft, Onyx, Wasatch,

                    Caldera,

                    Cheetah or any rip that I have ever used.

                     

                    To respond to your point very directly: I have established optimum ink

                    restrictions, ink limits and linearization. I then can output colors that

                    are achievable prior to the ICC workflow. Once the ICC profile is in place

                    often I run charts with and without vector CM. The result is that colors

                    untouched by ICC, while reducing control somewhat are far more vibrant. The

                    down side is we do not have the benefit and control of changing black start

                    and transitions of Light and Dark inks.

                     

                    Digital prints made on for example native 50DPI Spectra heads will deliver a

                    far more appealing apparent resolution with ICC profiles than without.

                     

                    What I would really like to see is some acknowledgment in books like Real

                    World Color Management that this is in fact taking place. Another good

                    example is the widely recognized scum or garbage dot that the ICC profile

                    introduces to remap a pure yellow to the absolute LAB values associated with

                    Y.

                     

                    Some of these problems are so common I am actually at a client site looking

                    at examples at this moment. It is like the Emperor's New Clothes. Many

                    people in our community are in denial that there is an issue.

                     

                    Thanks for your input,

                     

                    Steve Shaw

                    • 7. Re: ICC Profiles Clipping Gamut
                      bbeens Level 1

                      Steve-

                       

                      What software are you using to generate the icc profiles?  Also, the example you used of input 100% M being clipped to 70% M 3% C - does your rip report the source color as being out of gamut?  I would assume so as you state changing the rendering intent can produce better results.  I am surprised your icc profile is reducing ink so much.  I would start tweaking knobs in your profiling software and regenerate the .icc/.icm file to see if you get improved results.

                       

                      Not sure I follow with your comment regarding the 'garbage' or 'scum' dot.  I have profiled many devices produced by a dozen different companies using UV ink, solvent, aqueous and I have not seen the issue you are describing.  That is not to say there is no issue, I just haven't seen it.

                       

                      Any chance you are the Steve Shaw that worked at Gandi?

                       

                      Bryan

                      • 8. Re: ICC Profiles Clipping Gamut
                        PhotoButler Level 1

                        Hello Bryan,

                         

                        Your name is so familiar. Yes I am Steve Shaw who works at Gandi.

                         

                        Where do I know you from? By garbage or scum dot I am referring to the dot,

                        usually cyan or magenta that we see in for example yellow.

                         

                        Many clients insist on bypassing ICC for the reason that any impurity in

                        particularly yellow is visually disturbing. And yes the color was indeed out

                        of gamut. The particular Aquajet inks that i was profiling tend to over

                        saturate the media so it was a struggle to achieve a reasonable total ink

                        limit and still maintain reasonable dynamic range.

                         

                        Sometimes the scum dot is less evident depending on the particular printer.

                        On a high resolution printer if cyan is in yellow we just do not necessarily

                        see the dot. With the Jeti or Vutek I have consistently experienced this

                        phenom.

                         

                        The Colorburst rip even has a method for eliminating it. And finally I

                        tweaked all settings that I could possibly use to influence the result. The

                        profiling software I used on this occasion was Monaco.

                         

                        Have a good weekend, and thanks for the input.

                         

                        Steve

                        • 9. Re: ICC Profiles Clipping Gamut
                          PhotoButler Level 1

                          Hello Bryan,

                           

                          Now I remember: I am an idiot. How are you? I assume you are still with

                          Wasatch.

                           

                          Steve

                          • 10. Re: ICC Profiles Clipping Gamut
                            bbeens Level 1

                            Hi Steve-

                             

                            Yes, I am still with Wasatch.  Unfortunately I think this is a question for the good people at X-rite.  I am rather surprised with the example you stated above.  As to your original question, it seems a bit tautologicalThe ICC profile will clip the gamut to what it determines the devices max gamut.  But I think I understand where you are going.  Correct me if wrong - in your linearization chart you are seeing more saturated 100% patches than what your test images are showing?  If so, I would point you to check the working spaces for each of the files you are using.  Linearization charts often are created in Device space.  In some rips (Softrip for sure, not certain about Colorburst) this will bypass some color management tranforms but will go through linearization and ink restrictions / reductions.  The ICC profile is not going to affect this file.

                             

                            That is a complete guess due to my lack of knowledge of other rips.

                             

                            As for the scum dot issue.  The easy answer would be to train the designers to use LAB values for pure hues.  Have them spot capture a patch of 100% Y, C, M or whatever then use these LAB values when designing.  Or check the rip for a 'Spot color replacement' type control such as Softrip has.  Replace to Device space as mentioned above and the ICC won't monkey with your values.  ie 100% Y will not be mixed with C or M.

                             

                            Hope this helps.

                             

                            Bryan-

                            • 11. Re: ICC Profiles Clipping Gamut
                              Guillaume Muller

                              Steve,

                               

                              I've read this thread with interest, since I'm also very familiar with this problem - or more exactly my customers are familiar with it (I'm from Caldera).

                               

                              Bryan has mentionned one of the possible reasons you might have a "reduced" output gamut, it's when the source gamut itself is very small. But most of the tme, the problem is just the opposite: the printer gamut is smaller than the source gamut, so the profiler (the application that creates the profile) has to make choices when mapping out-of-gamut colors. Sometimes the result is not "optimum" at your eyes, in the sense that you would have prefered another choice, e.g. 100M instead of 70M+3C in your case. Another common example is when you print a pure red (100M+100Y) and it comes out as a kind of brown, whereas just 100M+100Y (i.e.: what you get without ICC) produces what you think a better red. This happens because the profiler has decided for some reasons  that this brown was "closer" to the input color than 100M+100Y. Note that in some cases the result will vary considerabily depending on the profiling engine you use, and even on the options of a specific engine. For example, if the brown and the pure red are both exactly at the same distance than the input color, and there is no other possible choise, then software A may choose the brown, and software B the pure red. That may be rather random.

                               

                              Now, that's relatively obvious, but they may be a third possible explanation: sometimes the color is in the gamut of the printer, but the profiler still will not select it, even in Colorimetric mode. Why? Because there is a difference between the "potential" gamut, which is the set of all colors printed on your ICC target, and the "achieved" gamut, which is the set of all colors that can be produced by the profile. The difference comes mainly from the Black Generation options you select, of course, and especially the Ink Max value. So mainly the shadows are affected. But other colors may be affected, because the profiler also has continuity constraints that can sometimes prevent it to reach a specific area in the gamut. That means that if you have 2 colors close in LAB (say L1 and L2) that can be printed optimally with 2 very different combinations of ink (say P1 and P2), usually the profiler will *not* map L1 to P1 and L2 to P2. Otherwise if you print a gradation from L1 to L2, you will see a kind of break coming from an abrupt variation of ink percentages on a short range. So the profiler will have to make a trade-off between continuity (or smoothness if you prefer) and color fidelity. And you will loose some colors in the achieved gamut (note also that in the case of an out-of-gamut color, as described in the first paragraph, this continuity constraint may play a role in the choice of the color. It's not just which of the brown and the pure red is closer to the gamut).

                               

                              So how do you prevent against that? Well, actually you can't. That will always happen, because mapping LAB -> CMYK (or more generally LAB->N-color) is all about this trade-off between smoothness and accuracy (I would even say that there is not other difficulty, but it's a big one). However, there are things that will contribute to create problems. For example bad linearization (and this includes bad transitions between light and dark inks, and bad transitions between the different drop sizes, in the case of a variable-drop printer), and global ink limitation too low (compared to the individual ink restrictions), or too complex. Everything producing discontinuities when printing the ICC target is likely to confuse the profiler when it has to evaluate the smoothness.

                               

                              Another question you might ask is how do I know if my profile missed a particular color? Well, in some Rips you have a tool to compute an "Optimized CMYK" combination corresponding to a particular LAB value. It's optimized in the sense that instead of using the profile, the Rip will look back in the original measurements of your ICC target to find the better match (somehow doing the same job than the profiler, but without continuity constraints). This usually can be found in the Spot Color or Color Substitution window. However that still doesn't tell you how far the achieved gamut of your profile is from the potential gamut, from a global point of view. I would say here that all RIPs are currently missing some features there. They should be able to display that kind of information after the profiling process (that would be helpfull were it only for the shadows). Ideally, they should even be able to compare the potential gamut with the "initial" gamut, i.e. the gamut you have before ink restrictions.

                               

                              Now, as for saying why "many people in our community are in denial that there is an issue", I can't answer. It's true that when you are looking for information about profiling, in books, paper, software documentation, etc.. you always find the same theoritical blah-blah about independant color spaces, the different renderings, and all that. It's true also that if you are a customer and call the tech support of your Rip software to ask why your red is brown, chances are that you will be answered that "the color is not in the gamut and that's all". At best you will be instructed to switch from percepual to colorimetric or to saturation (usually to find out that it's even worse than in perceptual). IMHO there is not one issue, but a combination of three here. First, some people do not understand that ICC is about color reproduction, and not about making the colors of their dreams. That makes things difficult from the start. Second, the profiling engines should be better in mapping out-of-gamut colors, or at least have more options. I haven't seen any major improvement in the 4 or 5 last years in the 2 main (commercial) profiling engines. And third, RIPs or software that use these engines should help the users to do better. Currently, to my knowledge none of the RIPs do anything else than computing a profile. You discover the problems only when you print, and then you have no hint as to how to solve them.

                               

                              As for the garbage in the yellow, I think that has nothing to do with this problem. My opinion is that there is nothing to do, in pure ICC. The only solution would be to compute a DeviceLink profile for each combination of profiles input/output, optimized to map pure yellow to pure yellow (or equivalently, to recompute an output profile "on-the-fly" for each different input profile). But of course if you do that, you don't talk anymore of color reproduction. Yellows are very differents from one printer to another, and from one standard to another (e.g. Swop vs Euroscale). From my point of view, the correct approach to solving this problem is the one that Bryan mentions: Spot color replacement or Color Substitution.

                               

                              Guillaume