Philip Sapirstein
I present new results in the digital analysis of inscriptions through a case study focused on IG XIV 1, the early sixth-century BCE text carved into the stylobate of the temple of Apollo at Syracuse. Over the past 150 years, scholars attempting a restoration have been challenged by the text’s unusual letterforms and location—on the first extant Doric temple in Sicily—and the lack of formulas found in later dedicatory inscriptions (Oliverio 1933; Guarducci 1949; 1982; and Svenson-Evers 1996 no. 42 compile the full bibliography). During fieldwork in 2015, I created 3D models of the inscription using photogrammetry. The digitally enhanced imagery of IG XIV 1 has made clear that we must fundamentally change the current readings of the text. First, it is necessary to restore another name for the dedicator; second, we must change our understanding of what he contributed to the temple; third, it is certain that the text must have continued well beyond the position on the stylobate where it has universally been thought to end. While this final conclusion excludes a definitive reading of the whole text, the revisions to the extant sections can be shown to specify a technical contribution, as if the dedicator were responsible for the innovative construction engineering of this early temple. In the paper, I discuss both the historical implications of the reading itself and the digital methodologies that led to this discovery.
The potential for digital enhancement of ancient incised surfaces has attracted much attention over the past decade (Orlandi et al. 2014). A common approach known as Reflective Transformation Imagery (RTI) compares multiple photographs, allowing the user to simulate artificial illumination and surface properties that reveal features difficult to detect by the naked eye (Malzbender et al. 2001; Mudge et al. 2005). However, RTI requires the illumination to be controlled throughout photography, which is a significant limitation when working at large sites outdoors, as with the Syracuse inscription. Full 3D recording is more suited to these conditions, typically employing some kind of structured-light scanner to record data (e.g., Anderson and Levoy 2002; Papadaki et al. 2015; or 3D from flatbed scans of squeezes: Bozia et al. 2014). The arrival of new photogrammetric software techniques known as Image-Based Modeling (IBM) now allows researchers to produce 3D models from no more than photographs (Remondino et al. 2014). The relatively low-cost method is capable of generating 3D models with more than sufficient resolution to restore nearly invisible incisions in a wide range of contexts, including rock art and inscriptions (Pires et al. 2015; Carrero-Pazos et al. 2016; Porter et al. 2016).
The analysis of IG XIV 1 has involved the development of new software designed specifically for common problems posed by Greco-Roman inscriptions. There are two key contributions which substantially improve upon previous work with RTI and IBM modeling. First, the code operates directly on the point clouds generated by IBM, rather than the sub-sampled 3D mesh, which allows the data to be visualized at much higher resolutions and quality than attained by other methods. Second, the technique restores the original surface into which the inscription was cut while accommodating for the presence of breakage, resulting in a more accurate assessment of the incision depth. Accordingly, random breaks can be filtered out from the enhanced imagery, and faint traces in abraded areas at the expected incision depth become more legible. This latter analysis of the plane of the inscription was responsible for revealing that the Syracuse text does not terminate at the last extant letter but instead continues, its final words erased by later trimming of the stylobate. I conclude with a brief discussion of how these techniques might be employed toward any number of other epigraphical problems.