I have seen dE below 0.1 average, but that is very hard to get because you (as Hal says) have:
- Measurement instrument variability - Plot-to-plot variability, even if you print patches, then profile and then measure. Good printers minimize that. - Printer-to-printer variability. Different printers behave different, same brand and model may have 4-5 dE easely. - Media. Unless you use roll, sometimes there may be differences from sheet to sheet, even in expensive glossy photo paper. - Workflow: THIS is the main source of errors. Double check the workflow. For a decent profiling program/media/printer you should have about 3-5 dE on average. Less that that is difficult, but can be done. More than that is probably a workflow issue. Make sure you use absolute colorimetric intent, make sure you use D50 illuminant. Use same instrument for profile and check if possible. Generic profiles are not accurate, expect 10dE if you are using a profile not built for your specific printer. And finally, make sure the color is in gamut, as it happens many times big dE are just on unrealizable colors. Regards Marti On Wednesday 08 April 2009 10:59:36 am Campbell, Jason J wrote: > One other question... If I give a CMYK input to obtain Lab output, and I > use a known CMYK set from the data used to create the profile, should the > returned value match the actual/measured Lab value? I am getting close > results, but they are far enough apart for me to question this... > > Example: > > Input CMYK> 0/100/20/0 > Measured Lab> 48.26/72.44/7.68 > LCMS Lab> 52.18/76.42/9.71 > > That's pretty 'off' for a value that was actually used in composing the A > side of the profile isn't it?! You need to remember that the measured data is very sparse and can be very noisy. Sources of noise include variation in the devices response over time and in different areas of the devices reproduction area (IE. different parts of the page if it is a printer) as well as variations in the measurement devices output. There are many other source of measurement noise as well. As a result the profiling software has to try to fit the curves used in the profile to that data in such a way that the noise that is present does not cause artifacts (IE. the curves are smoothed to minimize the influence of localized noise) and the fit to the measurement data will NEVER be perfect and in some locations can be fairly far off. In other locations the same type of comparison may yield a much closer match if the smoothed curves in the profile happen to fall close to the actual device measurement. In addition you didn't tell us how you derived the LCMS Lab value since things like rendering intent can cause these values to change significantly. Hal ------------------------------------------------------------------------------ This SF.net email is sponsored by: High Quality Requirements in a Collaborative Environment. Download a free trial of Rational Requirements Composer Now! http://p.sf.net/sfu/www-ibm-com _______________________________________________ Lcms-user mailing list Lcms-user@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/lcms-user
