Pigment identification by analytical imaging using multispectral images

This study utilized multispectral images to analyze pigments on a painting. This was accomplished by capturing multispectral images using a high-resolution image scanning system equipped with a monochromatic line CCD. The images were used to reconstruct spectral reflectance by solving a linear transfer function model representing camera response and material response. The CCD sensor response was filtered using Akaike Information Criterion (AIC) to eliminate the effect of noisy data. The pigments on the painting were identified by comparing the reconstructed spectral reflectance to a database of widely used Japanese pigments. The first derivative of the spectral reflectance was calculated using the Savitzky-Golay method to further improve the matching. Experimental results have shown good agreement between the actual pigments and the estimated pigments. Further developments on the technique employed in this study may eventually provide a beneficial tool for the nondestructive and noninvasive investigation of artworks.