The color reproduction systems, such as a color copier and/or a color printer, are widely used to visualize any graphical information as recent progress in electrophotographic technique. However, some kinds of problems have occurred because the color reproduction theories for these systems are based on the densitometric color value that depend on the difference of physical characteristics of color reproduction systems and/or chemical characteristics of the color materials; the difference of the reproduced color between each color reproduction system. Therefore, it is necessary to adopt CIE L*a*b* value or other colorimetric values as a device-independent representation to reproduce color accurately. Furthermore, we require a method of transformation between device-dependent color representation and independent color representations. Recently, Irie et al. and Funahashi have mathematically proven that the three-layered artificial neural network can approximately realize the continuous mapping with any accuracy unless the number of unit in hidden layer of the network is limited.In this study, we propose the transformation method which realizes the nonlinear mapping from CIE L*a*b* value to CMY dot area size by using a multilayered artificial neural network with a back propagation (BP) learning algorithm. The transformation accuracy of the proposed method was evaluated in terms of the color difference between original color chips and the reproduced color chips which correspond to the output of the trained artificial neural network. We show that the ability of the nonlinear mapping of the neural network can provide a practical and an efficient transformation method for color representation.