ABSTRACT The prevailing color characterization model was established on the assumption of channel independence. However, modern displays did not typically exhibit good channel independence on color reproduction. With the emergence of modern displays, the computing performance of the existing characterization models is insufficient. In this study, we propose a subspace polynomial fitting model, constructing polynomial fitting relationships across one, two, and three channels respectively, based on the channel independence of the display to determine the division criterion of subspace (0, M, N, 255), which can form 64 combinations of training samples. The model outperformed existing models, with the maximum ∆E 00 color differences of the 777 tested combinations between the model predictions and the measurements across the seven displays ranging from 1.27 to 5.73. In addition, based on the subspace polynomial regression model, 4913 data sets of multilayer perceptron (MLP) neural networks can be constructed for reverse color characterization. Similarly, 777 tested combinations were calculated for seven displays, resulting in maximum color differences of 2.35 to 3.83 ∆E 00.
Read full abstract