A color selection method that considers subjective principles is becoming more and mover necessary in a computer-based color management system. However one from an; engineering perspective has received limited attention, Previously, we developed an approach to measure the degree of color harmony by the similarity of a pleasure-related ;function, CLD (color linguistic distribution), to a RHD (reference harmony distribution) an a one-dimensional image scale of EXCITING-CALM Following this approach, a new color selection system is presented in this article by modulating the CLD on the image scale. The procedure used is to encode an inharmonious image as a CLD, then it is modulated and referred to as a RHD, called an imagery specification. Finally, the specified CLD is decoded and reconstructed as a harmonious image in an operated color space. Four types of RHD are compared: three ideal fuzzy sets scaled to extremely calm, quite calm, and slightly calm, and the practical one measured from social trends. Far verification, the harmony grade of the decoded image can be measured based on the referred RHD. As a result, four times the harmonic degree of the decoded image is gained than the originally encoded one, For application purposes, the proposed system is suitable for applications related ro human subjectivity. (C) 2000 John Wiley & Sons, Inc.