Effect of data scaling on color device model fitting

Abstract

Output devices in print production can be characterized by different characterization methods. One commonly used method of color device characterization is least squares fitting. In essence, the least squares fitting is used to determine the coefficients of a predetermined polynomial, such that the sum of squared differences between the values predicted by the model and the empirical data is minimal. The choice of the polynomial order and the cross product terms which best describe the behavior of a certain device is not obvious. This paper is a part of a larger study which investigates the criteria in the measurement data which can be used for optimal model selection. The part of the study covered in this paper addresses the data over fitting problem. It is investigated by comparing the performance of models of different polynomial orders on two different domains.