Ing. Martin Sivák wrote:
We used a metal plate of similar white as the mug surface is (white), but of course it wasn't perfect match. We then measured the targets from the plate.
100% color accuracy is not required (we could measure the cups, but ouch.. that would be too many), but at least neutrals and skin tones should be acceptable by common customers.
There are a couple of approaches that might work. One would be to make an MPP profile from a smaller number of patches measured on the actual cups, and then transfer it to an ICC profile using fakeread. Another might be to use a small number of measurements from the cup with refine, to improve the profile in critical areas. You would have to figure out what your critical colors are by sampling your input space, then convert them to CMYK/RGB for the current profile, print the cup and measure it, and feed the "should be" vs. "measured" values into refine. A third way would be to use an MPP profile made from measuring the actual cups to tune the readings made from the metal plate using a tool I've written for this purpose called "retargti3". It's not part of the GNU Argyll release though, so if that approach appeals to you, you'll have to contact me directly. Graeme Gill.