Emission spectra estimation method based on dimensionality transform strategy for liquid crystal displays

Lower accuracy of spectral information estimation from RGB values in liquid crystal displays affects high-precision color reproduction and display evaluation based on spectra. To improve the precision of spectra estimation, we propose an emission spectra estimation method based on a dimensionality transform strategy. Some metric learning techniques and kernel tricks are used for spectra dimensionality transform. The redundancy of spectral signals and nonlinearity of the display system have been removed; namely, the color feature is optimized. Then, a support vector regression model is applied in this dimensionality-transformed space to predict the emission spectra. The proposed method is validated by comparing the International Commission on Illumination (CIE) 1976 (L*, u*, v*) uniform color space and DE 2000 color differences between the estimated and true spectra for three different displays. The results manifest that the accuracy of spectra estimation obviously improves after applying the dimensionality transform strategy. The proposed method is significant for display color management, high-fidelity color reproduction, and display evaluation.