Volumetric Fast Fourier Convolution for Detecting Ink on the Carbonized Herculaneum Papyri

A novel approach using modified Fast Fourier Convolution (FFC) for volumetric data has demonstrated remarkable efficacy in detecting ink on the carbonized Herculaneum papyri, a collection of highly damaged ancient documents. This method, integrated into a deep learning-based segmentation architecture, significantly advances Digital Document Restoration (DDR) techniques by enabling more accurate and automated analysis of fragile artifacts. The research leverages X-ray tomography data to virtually unwrap and interpret the scrolls, addressing longstanding challenges in preserving and studying these irreplaceable cultural treasures. The proposed operator's versatility is further highlighted by its potential applicability to other volumetric data tasks, with implementation details made publicly available to foster broader scientific exploration.

Quattrini, F., Pippi, V., Cascianelli, S., & Cucchiara, R. (2023). Volumetric Fast Fourier Convolution for Detecting Ink on the Carbonized Herculaneum Papyri. 2023 IEEE/CVF International Conference on Computer Vision Workshops (ICCVW), 1718–1726. https://doi.org/10.1109/iccvw60793.2023.00188