Articolo in rivista, 2007, ENG, 10.1007/s10032-006-0028-7
Salerno E.; Tonazzini A.; Bedini L.
CNR-ISTI, Pisa, Italy; CNR-ISTI, Pisa, Italy; CNR-ISTI, Pisa, Italy
This paper reports some of the results obtained by applying statistical processing techniques to multispectral images of the Archimedes palimpsest. We focused on the possibilities of extracting the faint and highly degraded underwritten text,which constitutes the most ancient source for several treatises by Archimedes. Assuming each image to be generated by a linear mixture of different patterns, characterized by different emissivity spectra, the specific difficulty in separating the underwriting is that the mixture coefficients are unknown. To solve this problem, we rely on statistical techniques that maximize the information content of the processed images. In particular, we assessed the performances of the principal component analysis (PCA) and the independent component analysis (ICA) techniques. On the basis of 14 hyperspectral views of part of the palimpsest, we succeeded to extract clean maps of the primary Archimedes text, the overwritten text, and the mold pattern present in the pages. This goal was not reached in all the cases, because of the nonperfect adherence of the data model to reality. In most cases, however, PCA and ICA produced a significant enhancement of the underwritten text.
International journal on document analysis and recognition (Print) 9 (2-4), pp. 79–87
I.4.6 Segmentation, Blind source separation
Bedini Luigi, Salerno Emanuele, Tonazzini Anna
ISTI – Istituto di scienza e tecnologie dell'informazione "Alessandro Faedo"
ID: 68350
Year: 2007
Type: Articolo in rivista
Creation: 2009-06-16 00:00:00.000
Last update: 2018-01-24 09:39:08.000
CNR authors
External IDs
CNR OAI-PMH: oai:it.cnr:prodotti:68350
DOI: 10.1007/s10032-006-0028-7
ISI Web of Science (WOS): 000247735100002
Scopus: 2-s2.0-34147124519