Due to the dyeing process, learning samples used for color prediction of pre-colored fiber blends should be re-prepared once the batches of the fiber change. The preparation of the sample is time-consuming and leads to manpower and material waste. The two-constant Kubelka-Munk theory is selected in this article to investigate the feasibility to minimize and optimize the learning samples for the theory since it has the highest prediction accuracy and moderate learning sample size requirement among all the color prediction models. Results show that two samples, namely, a masstone obtained by 100% pre-colored fiber and a tint mixed by 40% pre-colored fiber and 60% white fiber, are enough to determine the absorption and scattering coefficients of a pre-colored fiber. In addition, the optimal sample for the single-constant Kubelka-Munk theory is also explored.
This paper is committed to improving surface color prediction accuracy of the single-constant Kubelka–Munk (KM) model for pre-colored fiber blends without increasing the model complexity. The single-constant KM model is only applicable to certain media with a constant scattering coefficient. However, the scattered lights in pre-colored fiber blends are intertwined with a great deal of fiber surface reflections, making it impossible to obtain the true KM scattering coefficient. To solve this problem, we analyzed the propagation behavior of light beams within the pre-colored fiber blends, and proposed a light scattering correction equation to remove the effects of fiber surface reflections on the scattered lights. Then, an improved single-constant KM model was established based on the corrected spectral data. Pre-colored cotton fiber blended samples were prepared to assess the color prediction accuracy. The results show the improved model, with coefficients k1 = 0.9477 and k2 = 0.0523, achieved superior performance compared to the original single-constant KM model and the two-constant KM model. The average color difference (ΔE2000) of the improved model is 1.20, while the average ΔE2000 of the original single-constant KM model is 6.37, and that of the two-constant KM model is 1.58. Importantly, the improved model has not added complexity to the single-constant KM model since the light scattering correction equation is essentially used to pre-process the spectral data. It can be concluded that the improved model is beneficial and practical.
Many ancient wall paintings are confronted with the threat of irreversible damages and in urgent requirement of restoration. This work provides the superpixel segmentation method and pigment identification method for the visible spectral image of ancient wall paintings to guide the scientific restoration of the paintings. The superpixel segmentation method for the visible spectral image is an extension of SLIC (Simple Linear Iterative Clustering) for the RGB image by redefining the feature of the visible spectral image. It can extract the outline of wall paintings and limit the pigment filling area in restoration of wall paintings. 44 kinds of commonly used pigments with size variations are selected to construct a visible spectral reference database for pigment identification. The pigment used in each superpixel is identified by searching the database in a specifically constructed feature space to find the nearest reference sample. This can provide guidance to pigment selection in restoration of wall paintings. At last, the methods are validated using the visible spectral image captured from Mogao Grottoes in Dunhuang by using a multispectral imaging system.