Press Characterization Recalibration by Principal Component Analysis and DEBP Model

摘要

Artificial neural networks (ANN) combined with PCA are widely being used. This study addresses the problem of updating a CMYK printer characterization in response to systematic changes (printing material, press) in device characteristics with PCA-DEBP model. In this study, training samples of test chart, which were normalized through principal component analysis (PCA), were applied as inputs to a differential evolution back propagation (DEBP) neural network with one hidden layer. This DEBP model has been used to predict the present printing characterization using the last ICC profile by measurement with high convergence speed. Experiment results show that the predicted printing characterization compares with that by the measurement have little color difference. So a PCA-DEBP model can be used to exactly recalibrate the ICC profile over time with low cost.