Color segmentation algorithms to support automatism of graphic documentation in restoration

Amura A., Tonazzini A., Salerno E., Pagnotta S., Palleschi V.

Cultural heritage  Physical restoration  Segmentation  Image processing  Image analysis  Neural networks 

All the disciplines connected to restoration have always shown a deep need to achieve a wide understanding of the object of interest, in order to develop a greater awareness of the actions needed and to perform in the full respect of its features. Diagnostic investigation and graphic design are the main tools the restorer operator uses to turn this dialogue into a process of representation where the design acts like a precise language responsible for synthesizing and plainly describing all the information. Our methodology starts with multispectral imaging coupled with segmentation algorithms for the automatic extraction of regions of interest (ROI) from raster images. Thanks to their different optical behavior, we can isolate different areas of the painting that are attributable to it's executive and restoration history. These areas, each corresponding to specific information, are then used for the representation and description of the edge and the area of the regions, producing outputs with vector attributes derived from the raster images. The methodology described above has been applied on a canvas depicting queen Cleopatra, attributed to Donato Creti. The last restoring intervention executed on the artwork is from 2019, during which the restorers have discovered several pictorial remakes, some executed by the artist himself and others executed after his death by unknown artists or restorers.

Source: XV Conferenza del Colore, Macerata, Italy, 5-7 September, 2019