RIS analysis of Ukiyo-e prints

Figure 1. Print by Utagawa Toyokuni I, about 1821. MFA # 11.30311

Overview

Reflectance Imaging Spectroscopy (RIS), also known as hyperspectral imaging, is a technique that acquires thousands or even millions of reflectance spectra across the surface of an object, using a hyperspectral camera. Each pixel in the “image” captured by the camera contains a complete spectrum from a specific spatial location. It is a non-invasive method whose principal application in the cultural heritage field is in determining the distribution of specific colorants or simple mixtures of colorants across the surface of a work of art. Hyperspectral cameras usually operate in the visible range as well as into the infrared, out to 2500 nm in some types. The camera used in our research acquires spectra in the visible and near-infrared range, between 400-1000 nm.

A RIS Case Study

The RIS images were generated by the Spectral Angle Mapping (SAM) algorithm, using the spectrum from a specific area as the reference. The SAM images show, in shades of gray, regions whose spectra most closely match the reference spectrum. The areas that are white are the closest matches. The print shown in this study is Actor Arashi Kitsusaburô I (Rikan) as the Monkey Trainer (Sarumawashi) Yojirô, in the Horikawa Scene of the Play Oshun and Denbei (Oshun Denbei Horikawa no Dan), by Utagawa Toyokuni I, dating to about 1821. The reconstructed RGB image from the RIS datacube of the analyzed print section (FIg. 2) is replicated on the right side of each section for comparison.

Selected Images

Figure 2. Reconstructed RGB image from the RIS datacube of the middle part of the print.

Figure 3. SAM (425-800 nm), using yellow at back of monkey’s robe as reference. This yellow was identified as turmeric by EEM. RIS suggests that the same colorant was used in the rope attached to the monkey. There is a suggestion that yellow is also present in the pink lips of the man (although this is not visible in the image, given the brightness and contrast settings used for the display).

Figure 2. Reconstructed RGB image from the RIS datacube of the middle part of the print.

Figure 4. SAM (425-800 nm), using pink part of monkey’s face as reference. The colorant was identified as safflower by EEM. RIS suggests that this colorant was also used in the lips of the man and in the small pink rings within the monkey’s red costume.

Figure 2. Reconstructed RGB image from the RIS datacube of the middle part of the print.

Figure 5. SAM (425-800 nm), using red costume of the monkey as reference. Colorant was identified as madder by EEM.

Figure 2. Reconstructed RGB image from the RIS datacube of the middle part of the print.

Figure 6. SAM (425-800 nm), using the darker blue edge of the man’s robe, near his chest, as reference. FORS analysis of a dark blue “star” decoration on the edge of the fabric behind the man’s right foot identified the colorant as indigo. SAM indicates that all of the darkest blue parts of the man’s costume, including all of the stars in the blue and white striped fabric, are the same blue.

Figure 2. Reconstructed RGB image from the RIS datacube of the middle part of the print.

Figure 7. SAM (425-800 nm), using the lighter blue of the man’s sleeve as reference. FORS suggested that the pigment is indigo, obviously less concentrated than in the darker blue areas. SAM indicates that all of the lighter blue areas of his clothes contain this same colorant. The spectra are different enough from those of the darker blues so that there is essentially no overlap in the SAM maps of the two blues, although they are made from the same colorant. Possibly the slight color of the unpigmented paper substrate is enough to shift the spectra of the lighter blue (less pigmented) areas away from the spectra of the darker blue (more heavily pigmented) areas. Note that both of the “two” blues appear to be present in the lower edging around the monkey’s costume. One or “both” blues also seem to be present in the greenish-tinted stick in the man’s hand. FORS and EEM tentatively suggested a mixture of indigo and turmeric in the stick, but RIS gave no clear indication of a yellow colorant, although one must be present. The spectra from the RIS datacube resembled those of the lighter blue – spectra with reflectance in the 500-600 nm which is higher than what is expected for pure indigo. This is further indirect evidence for a reasonable conclusion that the green was made from a mixture of indigo and a yellow, the latter most likely turmeric, due to the low amount of fluorescence shown by the green.

Figure 2. Reconstructed RGB image from the RIS datacube of the middle part of the print.

Figure 8. SAM (425-800 nm) of the pale-yellow paper. EEM and XRF suggested that both iron oxide yellow and a flavonoid were present. EEM results for flavonoids resemble EEM results from the paper substrates of prints, although the distinctly yellowish tint of the paper may support the presence of a flavonoid-containing colorant in this instance. XRF showed a somewhat elevated iron content relative to the unpigmented paper, which is the basis for the suggestion that an iron oxide yellow is also present in the background. No evidence of yellow iron oxide was found in spectra from the RIS datacube.

Note that the “flesh tones” of the monkey’s face and the man, as well as the man’s colorless scarf do not contain these colorants. They probably represent the unpigmented paper substrate.

Conclusion

The value of RIS is that it can indicate where a specific colorant, which may have only been identified or hypothesized in one small area by one of the point analysis techniques, is present within the entire print. The points analyzed by XRF, FORS and EEM spectroscopy on a print are chosen to include all distinct colorants that appear by close visual examination to be present on the print, but the points may not always have included all colorants, and the points do not show the distribution of a colorant over the entire print.