Today’s color reproduction systems aim at producing without manual intervention color images, which are as close as possible to the original images. Color images shall be transferable from one device to the next and still look similar when shown on a display or printed on paper. In order to be able to create on a given printer images whose colors are as faithfully reproduced as possible, it is necessary to characterize that printer, that is, to specify for a given selection of paper, inks, and halftones, the color response it provides when asked to print with given amounts of inks. Once characterized, a printer can be inserted into a color reproduction workflow. There are many additional reasons for characterizing a printer. Printer manufacturers may, for example, want to dynamically characterize their printers in order to verify that the printer operation mode has not drifted away. Depending on the result of this characterization, they may also adapt printing parameters in order to ensure that, for given control values, the printed colors remain identical. Finally, thanks to color prediction models, the usage of the inks can be optimized. In modern 6, 9, or 12 ink printers, there are many possibilities of printing a given color. The color prediction model may help the software to decide, according to given optimization criteria, which set of inks to select in order to print a given color within a given context. Possible optimization criteria can be theminimization of ink usage, the low visibility of halftones, or the minimization of metamerism when observing a print under various light sources. Since more than 50 years, there have been attempts to create odels predicting the color of printed halftone images. In order to offer accurate predictions, such models need to consider