中文摘要：

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