In this paper, a new analytical approach based on smartphone digital-images analysis coupled to chemometrics is proposed to investigate the nature of different pigments of historical and cultural interest. To that end, wall painting replicas were prepared and painted with 9 different pigments covering three colors: yellow, blue, and green. The mockups were photographed with a smartphone, and the images calibrated using a reference color chart. Two approaches were tested: one based on commercial color charts, and other based on lab-made ad hoc charts. The former demonstrated to yield better results, and the calibrated CIELAB coordinates were used to build Support Vector Machines (SVM) as a classification method. The optimized SVMs were able to successfully classify each pigment based on corrected CIELAB color descriptors, with errors of prediction below 5 %. Overall, this paper proposes a fast and non-invasive analytical method based on image treatment and chemometrics to discriminate among historical pigments in wall painting replicas.
Read full abstract