Chromatic distortion is one of the main problems of color pattern recognition. I present a scheme for an image system with a TV camera as input device to reduce the effect of chromatic noise. Notation and practical data of Munsell sample chips are recorded through the keyboard and the camera used, respectively. The notation can be looked up rapidly with a neural network. The network performs the transformation from RGB to Munsell space. Another adaptive resonance theory (ART) network is responsible to the transformation from the Munsell space to any desired space. The division of the two transformations ensures high system flexibility.