| Abstract: | SUMMARIES. When inpainting, selecting appropriate pigments that achieve an invariant color match can be a challenge for conservators, as changes in illumination and observer can result in metamerism. A simplified instrumental-based color-matching system, consisting of a portable, small-aperture spectrophotometer and spreadsheet software, was developed to aid in pigment selection. By using the single-constant simplification of the Kubelka-Munk turbid media theory and restricting the spectral analysis to 420nm and above, it was possible to develop a pigment database without the usual requirement of analytical weighing and homogeneous dispersion. Multiple-linear regression, common to most spreadsheet software packages, was used to select a set of pigments most closely matching the spectral properties of the original work of art. The methodology was tested in the conservation treatment of Burnett Newman's Dionysins' and Sanford Robinson Gif ford's 'Siout. Egypt '. The principal advantages of this system are the ease of database development, relatively low cost, nondestructive analysis, and speed of defining an appropriate set of pigments, particularly in comparison to common trial-and-error visual methods or complex microscopic or chemical analyses.
CONCLUSION. Spectral matching, used in many color-manufacturing industries, has been applied successfully to the problem of pigment selection for inpainting. In order to make this method practical in a museum setting, the single-constant Kubelka-Munk simplification was used, pigments were not weighed, and white was eliminated from the statistical analysis. Once the pigment database has been created, pigment selection using multiple linear regression can be a routine activity in a conservation department. The essential equipment required is a small-aperture, portable spectrophotometer and spreadsheet software. Despite the practical simplifications in comparison to industrial usage, the technique was clearly a significant aid in the conservation treatment of Newman's Dionysius and Gifford's Siout. Egypt. Because damaged areas of Dionysius were a high-chroma green, someone well versed in spectral-based pigment identification could have identified phthalocyanine green. Even so, selecting the best yellow pigment would have been challenging; a considerable amount of time was saved compared to testing various yellow pigments with phthalocyanine green. XRF, often used in the National Gallery of Art, Washington for pigment identification, would not have provided information helpful for selecting an appropriate yellow pigment. Furthermore, a different yellow pigment other than Naples yellow deep would probably have required additional pigments to achieve an acceptable color match, further complicating the treatment. In the case of Siout, Egypt, the likelihood of selecting raw sienna, cadmium yellow light and ultramarine blue was very remote. Because the mint-green color was so light and low in chroma and essentially lacking in pigmentation, it could have been visually matched by thousands of potential combinations of pigments. In the typical trial-and-crror method utilized by conservators, it is likely that many more than three pigments would have ended up in the mixture. In addition, the suggested mixture for this inpainting was somewhat counter-intuitive. Experienced conservators might well have avoided ultramarine blue, for instance, since it is known to often shift towards purple in changing lighting conditions and in photography. Likewise, raw sienna has a reputation among some conservators as often yielding serious color shifts when utilized in inpainting light colors. The conservator treating Siout, Egypt (Michael Swicklik) does not include it in his day-to-day inpainting palette for this reason. More common earth pigments, such as burnt sienna, burnt umber and yellow ochre, would have resulted in greater metamerism. For this painting, it was critical to achieve a nearly exact color match because the damaged and surrounding sky area was nearly homogeneous in color. Any degree of mismatch would be visible. In fact, the primary motivation for undertaking this treatment was that the degree of metamerism in the previous repair was unacceptable to the curators of the Gifford. The utilization of this color-matching technique was probably the deciding factor between success and failure in this very challenging inpainting problem. Although this technique has been very successful to date, we can imagine paintings in which the statistical selection is confounded by complex optical properties of the painting such as texture and transparency. It is possible to use the general form of Kubelka-Munk theory to predict spectral reflectance of either transparent or translucent paint layers on an opaque ground; however, creating the pigment database is considerably more complex than described in this research. Furthermore, nonlinear optimization would need to be used rather than linear regression: this increases the difficulty in generating and evaluating statistical goodness metrics. It is expected that applying this methodology for works of art on paper and dyed textiles will be even more successful. The single-constant Kubelka-Munk simplification is justified on theoretical grounds because the scattering of the paper substrate or textiles is always much larger than the majority of colorants used for these materials. Preliminary analyses with artists' watercolors have been very promising [30]. It is clear that an extensive database of pigments used in inpainting is very important to develop, particularly a large number of earth pigments. Having an extensive database might have resulted in a set of pigments for inpainting the Shut. Egypt sky that was even less metameric. Creating such a database will be a future research activity. As a final note, it is very important to use the same predicted pigments during treatment. A replacement pigment that looks identical visually may have very different spectral properties. Two pigments with the same name and Color Index number can be vastly different. Although it is expected that earth-type pigments vary widely with their source, the large difference between the two 'cadmium yellow light' pigments was unexpected. |
|---|
