Gauguin's Nativity was used as a case study for this research project. The left image is the Recto (front) and the right image is the Verso (back).

To measure the 3D surface of Gauguin’s graphic works we used a well-established computer vision technique known as photometric stereo. We captured a sequence of photographs of the artwork from a fixed camera location while changing the direction of the light source. This allowed us to calculate the surface normal at each point by solving a system of linear equations from our set of measurements at each pixel. Our photometric stereo setup consists of a color checker for color calibration, a 3D calibration target for 3D surface calibration, a reflective sphere for calibrating light direction, and the work of art.

In this figure we separating surface color from shading information for the 3D color calibration target. (Left) A captured photo of the calibration target. (Right) A relit photo synthesized from recovered normals visualizing surface shading independent of surface color.

Here we compare the 3D reconstruction with the 3D calibration target. The height of the cone in the calibration target is approximately 10mm. The surface reconstruction of Nativity’s protruding lines are approximately 100× smaller, indicating they are on the order of 100μm in height.

Here we show surface shading information for regions in Nativity corresponding to transferred lines. The transferred lines show clear evidence of surface shapes that protrude from the page.

This figure shows several frames from an animation visualizing the 3D surface shape at the location of the lines drawn in Nativity. We zoom into the area with the transferred lines along with a few frames of a 3D animation. As we peel off the color and shift perspective to look at the 3D surface shape, we see clear evidence of protrusions on the page where ink has been deposited. This is solid evidence of the ink being transferred from a matrix such as that in a monotype transfer process.

In this figure, we show clearly that there are no surface features at the locations of the blind incisions. This is solid evidence that the blind incisions do not originate from any indentations in the paper surface during ink transfer.

Here we show a map of the blind incising. Based on our findings, we hypothesize that this map is actually an inverted impressions made from the inked support that Gauguin used in his studio more than a century ago. Based on our hypothesis, this map shows where ink was removed from the glass matrix during the first step of the transfer process (shown in the next figure).

A visual demonstration of the hypothesized nine-step transfer process. First, a standard monotype is made by placing a piece of paper on an inked surface (the glass matrix) and drawing on the back. The monotype produced is removed from the glass matrix and set aside. Now the glass matrix lacks ink in the locations where it was transferred to the first monotype. We then place a second piece of paper on the same inked surface; this piece of paper will eventually be our final transfer drawing. We then draw on the back of this sheet of paper to transfer the initial ink layer, which in this example, is brown. In the process of making the drawing, fingerprints and marks from the pressure of the artist’s hand on the paper, deliberately or inadvertently, are also transferred. When we lift our sheet of paper from the matrix, we see that we have transferred inked lines to the surface, but the lines are broken in locations where ink was previously removed from the matrix while creating the first monotype. We then place the same sheet of paper that now carries a brown ink impression onto a second glass matrix inked with darker pigment. This matrix is pristine. We again draw on the back of the sheet to transfer the darker lines that form the second layer of transferred media. When we remove the paper, the process is finished and we have created the final transfer drawing.

Visual comparison between Gauguin’s Nativity and the transfer drawing produced using our hypothesized transfer process. The marks on the reconstruction appear visually consistent with the original.