Through discussing the color matching technology and its application in printing industry the conventional approaches commonly used in color matching, and the difficulties in color matching, a nonlinear color matching model based on two step learning is established by finding a linear model by learning pure color data first and then a nonlinear modification model by learning mixed color data. Nonlinear multiple regression is used to fit the parameters of the modification model. Nonlinear modification function is discovered by BACON system by learning mixture data. Experiment results indicate that nonlinear color conversion by two step learning can further improve the accuracy when it is used for straightforward conversion from RGB to CMYK. An improved separation model based on GCR concept is proposed to solve the problem of gray balance and it can be used for three to four color conversion as well. The method proposed has better learning ability and faster printing speed than other historical approaches when it is applied to four color ink jet printing.
