Principal component analysis of visible and near-infrared multispectral images of works of art

Principal component analysis (PCA) was applied to a very simple case of a tempera panel painted with four known pigments (cinnabar, malachite, yellow ochre and chromium oxide). The four pigments were spread pure as well as dilute with carbon black (5% w/w, 10% w/w) thus creating 12 homogeneous areas of the same size. The panel was imaged by a Vidicon camera in the visible and near-infrared regions (420–1550 nm) resulting in a set of 29 images. PCA was applied by taking various subsets of the input data. From the analysis of this simple and predictable case study some guidelines are synthesized and proposed for the application to actual work of art. Results are presented for the painted panel. Preliminary results are also reported for the Luca Signorelliʹs “Predella della Trinità”. The multivariate image analysis results in the visible and near-infrared regions show that it is possible to use the multispectral image data in order to get a segmentation and a classification of painted zones by pigments with different chemical composition or physical properties.