Specifications

Submitted to Studies in Conservation, March 2006
13
illumination geometry, the calibration target spectrophotometric measurement geometry, and
differences in gloss between the calibration targets and the painting. An interesting result was
that the correlation spectra were similar in shape for both systems. The spectral performances
were not statistically significantly different.
Conclusions
A practical spectral-based imaging system has been developed in which a color filter array
(CFA) digital camera was combined with two absorption filters. By taking two sequential
images, one with each filter, and deriving a linear transformation matrix using a target of colored
patches with reference spectral reflectance factor values, spectral images were estimated. The
performance was evaluated for test targets and an oil painting, Henri Matisse’s Pot of
Geraniums. This system had spectral and colorimetric performance that was equivalent to a true
spectral imaging system consisting of a monochrome camera and liquid-crystal tunable filter.
The main advantage of this new system is that any CFA professional-grade camera
system can be used, following a simple modification of replacing the detector cover glass and the
addition of a filter slider, or wheel, or holder (i.e., changing filters manually). This means that
scientific imaging of the visible spectrum can be the responsibility of an imaging department
rather than a highly specialized instrument requiring a conservation scientist well versed in color
and imaging science.
Because the system does not have a permanent infrared cut-off filter, it can also be used
for infrared reflectometry; the CFA has spectral sensitivity over a wavelength range similar to IR
film. Using an appropriate visible-spectrum cut-off or bandpass filter, the camera can be used as
a conventional RGB camera, a spectral camera, or an NIR camera.
Future research will focus on imaging and spectrophotometric geometry considerations,
improving the spectral accuracy for neutral samples, an improved calibration target, and an
Adobe Photoshop implementation of the Matlab software.