Palimpsest Research Based on Hyperspectral Information

A methodology for restoring palimpsests using a specialized hyperspectral scanner is proposed. The methodology includes the following analysis methods: principal components, independent components, unsupervised machine learning, video analytics. The analysis of the obtained results shows that out of 13 samples for 7 it was possible to reveal the hidden text that is not visible in the color image or with the naked eye. For 4 documents, we managed to increase the local contrast of the text of the second layer and make it more readable.