AIC's 44th Annual Meeting

In this talk we will show how reflectance transformation imaging (RTI) can be used as a quantitative technique capable of visualizing and measuring the surface shape of works of art. RTI utilizes multiple images captured from a fixed camera position but lit from various different directions to create an interactive composite image that reveals textural characteristics of materials. While current RTI methods offer conservators a powerful exploratory tool, the many systematic approximations inherent to the technique limit its use to qualitative assessments of appearance. As one step towards quantitative surface estimations, we address a fundamental limitation of the RTI model, that the whole object is lit from the same illumination angle with the same illumination intensity across the entire field of view. This requirement is rarely met in real-life experimental conditions because the light would need to be placed infinitely far away from the object. The mismatch between the lighting model and real experimental conditions has been documented to produce erroneous surface normal estimations, a “potato-chip” shape estimation error when the surface normals are integrated, and non-uniform illumination effects in relighting that we call the ‘spot-light’ effect. Using our new algorithm and capture methods, we show how to correct for these errors and advance RTI making it a practical and repeatable method to digitally capture the surface texture of a work of art. As a practical example to demonstrate the effectiveness of this approach, we will show high-quality 3D reconstructions of RTI data captured from the Art Institute of Chicago’s collection of the graphic works of Paul Gauguin.