Measurement of colour using a non-destructive method for the study of painted works of art

• Published in: Measurement science & technology 1994-09, Vol.5 (9), p.1078-1088
• Main Authors: VALLARI, M, CHRYSSOULAKIS, Y, CHASSERY, J. M
• Format: Article
• Language: English
• Subjects: Analytical chemistry; Chemistry; Exact sciences and technology; Miscellaneous
• ISSN: 0957-0233; 1361-6501

Art historians, archaeologists and also restorers of art works are called upon, within the scope of research and study of painted works of art, to determine and describe with accuracy the colours used by each artist. On the other hand, scientists who work on data banks of painted works of art are called upon to acquire very-high-resolution images combined with accurate colorimetric registration for archiving purposes. An extended study oriented towards colour measurement parameters determination is then necessary. From the current available relevant literature it is obvious that the research concerns primarily colorimetric fidelity without providing any details as far as mathematical processing of colour measurement data is concerned. In the present study there is suggested a non-destructive methodology for colour measurement that is much cheaper than the one that is applied if non-destructive spectrophotometry is used for this purpose. This is performed in a comparative manner with the help of a three-charge-coupled device or a mono-charge-coupled device colour detector five times less expensive than the first one, on two reference panels containing a total of 34 colour samples. Aiming at the uniquely and generally accepted determination of colour, the R, G and B values are first properly corrected and then transformed to x, y (chromaticity) and Y values in accordance with the CIE 1931 standard colorimetric XYZ system. This procedure is performed with a linear mathematical transformation followed by a least-squares method in order to determine the best values of the matrix elements and minimize the error between measured and computed values. The so-obtained values are then compared with the respective x, y and Y values provided by the spectrophotometer on the same reference panels. The mean relative difference between the chromaticity x and y values provided by the set of the two detectors and the values coming from the spectrophotometer is approximately 1.7%. The collected data expressed in tri-stimulus values are finally converted into CIELAB notation (L a b) for data evaluation. An average colour accuracy of about three units in the CMC uniform colour space was achieved. These results offer a very promising attempt at elaborating a protocol of colour measurement in a reproducible way.