39th Annual Meeting – Paintings Morning Session, Friday June 3 – “An Experimental Study on the Merits of ‘Virtual Cleaning’ of Paintings” by John Delaney

John Delaney discussed the “yellow filter” model for virtual cleaning and compared it to experimental results from measurements of cleaned paintings before and after treatment.  To estimate the result of removing a discolored varnish, conservators sometimes correct the color using imaging software, “subtracting” out the varnish as if it were a yellow filter.  The experimental values were collected using FORS (fiber optic reflectance spectrometry) of specific points and hyperspectral imaging of the entire surface of the paintings.  The specific values for the  “yellow filter”  for a given painting is taken from the absorbance spectrum of the removed varnish and then applied to the rest of the painting.  After examining the spectra before cleaning, after cleaning, and the estimated virtual cleaning, it was clear that the yellow filter model for virtual cleaning was not enough to explain the spectral changes seen after cleaning.  After varnishing the cleaned area, there was no change in color, only an increased brightness indicated by a rise in the near-infrared signal, so that alone did not explain the additional component missing from the yellow filter model.

I was thinking that even if it wasn’t spectrally perfect, could the yellow filter model at least be a good tool for an estimation of results?  The example of the 15th century Tuscan panel convinced me otherwise—while the yellow filter worked well for the blue areas, the model was a poor estimate for the red and brown areas.  One component that is not taken into account with the yellow filter model is the effect of scattering by the varnish.  Aged varnishes don’t just yellow, they also increase the scattering of light.  Even with a scattering model included, the yellow filter model still did not accurately reflect the results of removing a yellowed varnish.

While virtual cleaning with simple “yellow filter” color correction in image processing software such as Photoshop is cheap and can be somewhat informative, it certainly is not a perfect model for varnish removal, even when the yellow filter’s spectrum is taken from the actual discolored varnish.  The study has not considered selective cleaning, grime removal, or other nuances of cleaning paintings, which would only further complicate the spectral correction.