ABSTRACT Accurate estimation of the absorption ( K ) and scattering ( S ) coefficients in the two‐constant Kubelka–Munk (KM) model is essential for spectral reflectance prediction and color matching in precolored fiber blends. However, conventional least squares approaches often suffer from numerical instability and require large amounts of calibration data, limiting their practical use in real‐world manufacturing. This study presents an improved least squares method that incorporates a simple constraint by fixing the scattering coefficient of the undyed white fiber. This modification reduces the number of unknowns and improves the conditioning of the coefficient matrix, enabling accurate KM coefficient estimation even with limited calibration data. The method is evaluated on 48 blending samples of five pre‐colored cotton fibers using cross‐validation over a range of training sample sizes. Prediction performance is assessed using CIEDE2000 color difference (Δ E 00 ), spectral root mean square error ( RMSE ), and the condition number of the system matrix. Results show that the proposed method achieves lower prediction errors and improved numerical stability compared to the original formulation, even with fewer training samples. These advantages make the method readily applicable in industrial color spinning and sustainable textile coloration.
Li et al. (Mon,) studied this question.