Support vector machine-based method for colorimetric characterization of computer-controlled liquid crystal displays

Abstract

This paper proposes a colorimetric characterization method based on support vector machine (SVM) for computer-controlled liquid crystal displays (LCD). First, forward and inverse colorimetric characterization models based on SVM are established for LCD monitors. Second, the radial basis function (RBF) kernel function is introduced to deal with the nonlinear problem of color space conversion. Then, the parameters Gamma, penalty coefficient C, and ε are determined by 10-fold cross-validation grid search optimization. Next, the effectiveness of the method is validated by three brands of monitors; for the forward colorimetric characterization model, the average predicted color differences are 1.259, 1.091, and 1.005 CIELUV units, respectively; for the inverse colorimetric characterization model, the average goodness of fit are 0.985, 0.989, and 0.988, respectively. Finally, the method is compared with piecewise linear assuming variation in a chromaticity model, three-dimensional look-uptable, polynomial regression model, and artificial neural network model, and it is verified that the method is superior to the related colorimetric characterization models in color prediction accuracy. The method can better support the color management and high-fidelity reproduction of monitors.