Comparison of the Color Values Prediction by a Matrix Method and a Well-Known Software for a Digital Color Management Process

A condition that must be satisfied for digital printers in a color management system is converting device-dependent color values into device-independent color values by a transfer model of high-quality performance. Usually, each transfer model is initially trained with some specified samples as a training set before being tested. In the present study, color charts with different quantities of color patches in an RGB space were selected and printed as training groups.  The spectral reflectances of all samples in printed color charts were measured and converted to an XYZ color space under a standard observer/illuminant combination. Using a matrix method, the RGB to XYZ transformation matrices were obtained with the algebraic operation of different degrees of polynomials and the least-square errors for the training groups.Similar operations were also performed for the well-known ProfileMaker software, which is based on a look-up table method. Finally, the statistical values of color difference (DE2000) were calculated between the actual XYZ from the printed color charts, the predicted XYZ from the matrix method, and those predicted from the ProfileMaker software. The ProfileMaker software showed a significantly lower color difference near the perception threshold for the specified training samples.