Standard Finishing
4Over

Announcement

Collapse
No announcement yet.

Profiles generated with minimal number of patches

Collapse
Graph ExpoPrintPlanet
X
  • Filter
  • Time
  • Show
Clear All
new posts

  • Profiles generated with minimal number of patches

    I am not sure but there seems to be a trend.

    Here is a WTT article about GMG's OpenColor 2.1 which claims to use a very limited set of patches that can be used to develop a profile.

    http://whattheythink.com/news/85705-...t-test-charts/

    A new startup Canadian company, Color XTC also seems to have a similar product it is developing.

    I won't discuss the various problems with this approach but there could also be some positives.

    It could be interesting. A custom profile made with a very limited number of measured patches might be better than an industry generic profile that is sure to not accurately represent how a specific press prints.


  • #2
    Originally posted by Erik Nikkanen View Post
    It could be interesting. A custom profile made with a very limited number of measured patches might be better than an industry generic profile that is sure to not accurately represent how a specific press prints.
    To me that is the advantage, bringing presses that would not have been profiled due to time/perceived difficulty/etc into a profiled environment.

    Comment


    • #3
      OK. This is the deal as understand it. We currently use test charts with lots and lots of patches and then take LAB readings from them to build the profile.


      Now, when you take a spectro reading, about ninety percent of the data is discarded. It's not needed to create the LAB value.


      So GMG spent years scanning everything they could find and looking at the spectral data from the spectro reading. The maths used in open color makes the best use of the all the spectral data to predict the results on press.

      So you have less patches but the same, if not more data. I think open color uses a wedge of about eighty five patches but they have done tests with far less.

      I guess the final aim is to fit everything in a control strip along the side of every sheet.

      I'm sure if you contact your GMG rep they can do a proof of concept for you.

      Comment


      • #4
        Originally posted by Tim-Ellis View Post
        Now, when you take a spectro reading, about ninety percent of the data is discarded. It's not needed to create the LAB value.
        Sorry, what does that mean? What data is discarded? In order to convert a spectrum to CIEXYZ and then to CIELab, you multiply the spectrum by the standard observer spectra and the illuminant, and compute an integral over the result. Nothing is discarded - everything is required to calculate accurate CIELab values from the measurement.

        I agree that a spectrum allows higher accuracy than CIELab values when predicting - as opposed to measuring - colors. Using spectral data with sophisticated color printing model equations (such as Neugebauer, Kubelka-Munk etc.) will enable you to build an accurate model of the behavior of the press with less patches. But even allowing for spectral data in ICC profiles instead of CIEXYZ or CIELab, a lot of color matching problems arise from the design of the ICC profile format itself. Hopefullu the new iccMAX specification will fix some of these limitations.

        Comment


        • #5
          Originally posted by Schnitzel View Post

          Sorry, what does that mean? ....
          Hello Schnitzel,

          Well, with hindsight, perhaps that is a question I really should have asked myself before I posted it.

          The comment came from Juergen Seitz during a lecture I attended (he's GMG's Senior Technical Advisor and a German delegate in TC130- WG3.)

          At the time I took it at face value. I think it might just be a throw away reference to the string of thirty or so numbers in the raw data from the Spectro being reduced to three numbers for the LAB value.

          Anyway - I've reached out to him for an answer and if he gets back to me I'll let you know.



          Comment


          • #6
            Thanks, I'd like to hear more about it. I'm quite interested in GMG's approach to color management.

            Comment


            • #7
              Peter Schoeffler, Product Manager at GMG, comments: “For color-accurate proofing of spot color overprints, GMG OpenColor is the only available solution”.
              Which is of course not true.

              GMG’s main competitor, CGS also have a spectral data based system for measuring and predicting spot colour overprinting behaviour. This is available as a stand alone toolset and is also leveraged in products such as CGS ORIS Color Tuner and ORIS Flex Pack for inkjet proofing. The CGS approach uses an open ISO standards based approach, rather than proprietary/closed.

              http://color.org/CxF_test.xalter

              http://www.cgs-oris.com/en/products/oris-cxf-tools


              Stephen Marsh
              Last edited by Stephen Marsh; 06-17-2017, 07:34 PM.
              Comments are personal and my views may not be shared by my employer or partners.

              Comment


              • #8
                Originally posted by Schnitzel View Post
                Thanks, I'd like to hear more about it. I'm quite interested in GMG's approach to color management.
                Hello Schnitzel,

                i got a reply this morning. Here is my question and the response for you:


                Hello Juergen.

                I came to the Seitz Session in London back in 2015 (I think). Either you or Tony said that when we take a spectro reading about 90% of the raw data is discarded. Can I ask what is meant by that. Is it simply a string of thirty numbers is reduced down to just three?


                Hi Tim,

                yes, thats what i tried to tell. spectral readings are containing reflection values for the many different wavelengths. those all together would then describe a color under one specific viewing condition (eg D50) ... and this can then be described with an Lab-value

                but from the Lab value we could not estimate anything for another viewing condition (eg TL84 or D65, ...) only with the broader information from the spectral values, we could do such an estimate...

                Juergen Seitz


                I believe the approach in Open Color is to collect more data from less patches. The original Open Color 1.0 could use just a single 100% patch for each colour.


                http://www.worldpressonline.com/Pres...ard-39546.html



                In the lecture we were told a story where, armed with just 6 patches and the ink sequence, GMG were able to reverse engineer a 6 colour job that went wrong for a UK client after it was printed in India.


                Sadly I believe the approach is proprietary/closed. Open Color 1.0 was either very cheap, bundled with another product or given away free at shows. Once GMG began to realise its potential it then it was withdrawn and Open Color 2.0 was released.

                I hope that helps.

                Last edited by Tim-Ellis; 06-22-2017, 01:35 PM.

                Comment


                • #9
                  Thank you, Tim, for the follow-up. I see the movement from CIELab based color-management to spectral based as a good thing. An open standard would be better, of course.

                  Comment


                  • #10
                    Originally posted by Schnitzel View Post
                    Thank you, Tim, for the follow-up. I see the movement from CIELab based color-management to spectral based as a good thing. An open standard would be better, of course.
                    Which should be what ICC Max does.

                    Comment

                    UltimateDuploSmartsoft (Presswise)Standard FinishingKBA4Over
                    KBAKBA

                    What's Going On

                    Collapse

                    There are currently 4913 users online. 92 members and 4821 guests.

                    Most users ever online was 5,872 at 11:39 AM on 09-30-2016.

                    Working...
                    X