Demands on the spot color printing, which uses not only CMYK (Cyan, Magenta, Yellow, and blacK) but also several special inks, increase rapidly in the printing industries. Therefore, the quality of printed output and the efficient usage of various ink colors have to be improved in the competitive printing market. However, the spot color printing system usually depends on the experiences of the skilled workers for the elaborate output, and it would cause lots of manufacturing time, cost, and/or quality gap of prints. Therefore, we suggest a deep neural network model based on the datasets from the real-world printing system. The input data set consists of various characteristics of primary color inks and papers used to print, and the output data set is reflexibility of the products. Total 31 ranges of visible light are used to measure the reflexibility, and those values are converted into the standard color coordinate, CIELAB. The neural network model is tested on all the primary color space, and the average difference between the real CIELAB value and the predicted value is less than the 8.13, which means that the difference cannot be identified by untrained people. Further researches based on the controlled date aggregation using better spectroscopes would be essential to apply the suggested model to the real-world printing systems.
Read full abstract