Spectral-based characterization of inkjet printers by neural and genetic algorithms

The spectral-based characterization of inkjet printers is often based on a physical description of the printing process, but our experience tells us that it is not unusual to be unable to fit a mathematical model to a given printer. The objective of our work is to see whether the use of methods such as genetic algorithms and neural networks is an effective strategy to approach the complex problem of printer spectral modeling. In our experiments, we make no assumption concerning the sequence of inks during print and the printers are treated as RGB devices (the printer-driver operations are included in our model). To test our characterization methods, which require only 133 measurements for training the learning algorithms, we consider different inkjet printers and use different kinds of paper and printer drivers. The test set used for model evaluation was consisted of 777 samples, uniformly distributed in the RGB color space