AIC's 51st Annual Meeting

Between 2016 and 2021 a prominent project was undertaken to study and restore the large copy on canvas of Guilio Romano’s Battle on the Milvian Bridge by a Roman workshop ca. 1700 (now in the Court Bladelin, Bruges). In support to the restauration process of the painting the analytical research techniques and methods of scientific imaging, carried out by KIK-IRPA, IPARC and KU Leuven, brought to light new information at a very detailed level about the original composition of the pictorial layer and the later adaptations of the paintings. The MA-XRF analyses identified most of the original pigments and the overpainting during the various restoration phases in the nineteenth and twentieth centuries.
The condition before, during and after restoration was documented by the Multispectral Microdome (MSMD) based on photometric stereo computer vision algorithms. As a result, the spatial evolution of the restoration and the necessary improvements in the stability of the canvas and the pictorial layers could be studied and visualized in detail. This varying information obtained during the restoration was documented in detail for one selected crucial and challenging zone in particular: the eye of the horse on the scene of the battle on the Milvian Bridge.
The MSMD (photometric stereo, i.e. related to RTI) generated interactive imaging of the pictorial layers during different steps of the restoration. This tool illuminates a surface from 228 different directions with 5 different spectra: ultraviolet, blue, green, red and infrared. With each new lighting angle, a photo is taken, and with that complete dataset of images, computer algorithms can reconstruct both the relief and reflectivity of the pigments. The device thus documented five successive phases in the conservation-restoration process, each time the exact same zone. Phase 1 concerns the condition with varnish before the start of the restoration; phase 2 is after further decrease of the varnish layer; stage 3 is after the decrease of the fillers, making the cloth visible; in phase 4, the soil layer is brought up to level; phase 5 is after the completion of the restorations, with the new 'retouches' and the first new varnish layer.
This type of imaging makes it possible to interactively analyze the materiality of the pictorial layers. With the computer models, a researcher can virtually scan the surface in great detail. In addition, the incidence of light and light intensity can be controlled and the reconstructed information can be visualized in different ways. By working interactively with this, this research method supports the conservation-restoration process and ensures a rich documentation of the completed phases. In addition, this type of multiple rich datasets, including 3D-information, opens new possibilities for computer aided algorithms to perform automated comparisons between restoration phases. It also increases accuracy and reliability when final and intermediate results are compared where the physical differences between the restoration phases are severe. Standard 2D documentation methods often hit their limits here as too few points of references are detected between such phases to guarantee correct alignment.