Specifications

Submitted to Studies in Conservation, March 2006
8
integral shutter. This system was used routinely at NGA. The calibration target was a
GretagMacbeth ColorChecker SG. Commercial software was used to create an ICC camera
profile for the NGA system. The MCSL-Sinar and NGA-Sinar systems used the identical
lighting set-up, a pair of Broncolor Xenon strobes, positioned at approximately a 70° angle from
the surface normal on either side of the artwork. One side had about twice the irradiance. This
resulted in a quite collimated, raking illumination.
For this experiment, four targets were evaluated, described in Table 2. The
GretagMacbeth targets were available commercially. The Gamblin target was produced by
mixing each of the 31 conservation colors with titanium white at two different concentrations
and applying them to canvas board using a brush. The Blues target was made by mixing Liquitex
Artist Acrylic ultramarine, cobalt, Prussian, and phthalocyanine blue paints and titanium white in
various proportions in order to create a target with a large range of spectral characteristics. These
paints were also applied to a canvas board using a brush. The experiment consisted of imaging
the four targets followed by processing as described in Figure 1 and Eqs. (1) – (5). The targets
were used as either calibration or verification targets.
The colorimetric performance of the three systems is shown in Table 3. Both ∆E
00
and
∆E*
ab
are shown, the former metric having better correlation with subjective evaluations of
adjacent uniform color fields [16] and the latter having better correlation with images [17].
Experientially, the 90
th
percentile results are a better indicator of performance than the maximum
error, particularly for actual experiments (rather than computational analyses).
The Best Case corresponds to the MCSL-Sinar system where the same target was used
for calibration and verification. This provides a sense of the best mean performance that can be
achieved using the MCSL-Sinar system and the linear signal processing workflow as described.
These results are excellent indicating that the multi-filter system is capable of high color
accuracy. The small differences in performance between matched calibration and verification
targets (Best Case) and independent calibration and verification targets (MCSL-Sinar) indicate
that the choice of calibration target is important. Although progress has been made [18, 19],