Quantitative analysis of mixed pigments for Chinese paintings using the improved method of ratio spectra derivative spectrophotometry based on mode

Abstract

Abstract Background Mixed pigment analysis is an important and complex subject in preserving and restoring Chinese paintings since the colors observed by our naked eyes or instruments such as hyperspectral cameras are usually a mixture of several kinds of pigments. The purpose of this study was to explore a more effective method to confirm the type of every pure pigment and their proportion in pigment mixtures on the surface of paintings. Methods Two endmember extraction algorithms were adopted to identify the types of pigments and an improved method of ratio spectra derivative spectrophotometry was used to determine their proportion. Main works (1) Extracting the pure pigment components from mixed spectrum by adopting two blind source separation algorithms: Independent Component Analysis and Non-negative Matrix Factorization; (2) matching the separated pure spectrum with the pigment spectral library built in our laboratory to determine the pigment type; and (3) calculating the proportions of mixed pigments using the Ratio Spectra Derivative Spectrophotometry based on Mode, which is improved based on the original algorithm. Finally, a comparison was made with two abundance inversion algorithms: Least Squares Algorithm and Minimum Volume Simplex Analysis. The correlation coefficient and root mean square error were used to provide evidence for accuracy evaluation. Conclusions (1) Non-negative matrix factorization is more suitable for endmember extraction; and (2) Ratio spectra derivative spectrophotometry based on mode is more suitable for abundance inversion.