Prediction of colorimetric measurements in newspaper printing using neural networks.

For an optimum quality control in newspaper printing, patches printed with different combinations of CMYK should be colorimetrically measured. However, control strips with a large number of color patches cannot be included into the layout of newspapers, due to the additional space they need. Moreover, such control strips require a considerable amount of time for the measurement. To overcome these problems, it is preferable to print and to analyze only a few color patches and to obtain most information out of these. A method using neural networks has been developed to predict color values of two and three-color overprints from those of the primary inks, as well as color values of the primary inks from the two and three-color overprints. The feature of the method is that spectra are predicted from which colorimetric values (XYZ and L * a * b * ) and densitometric values are derived. The accuracy of the predicted CIELab values of the primaries from those of the overprints is typically within ΔE * ab = I for the two-color overprints, and within ΔE * ab = 2 for the three-color overprints.