TY - JOUR
T1 - Nonlinear Unmixing of Hyperspectral Datasets for the Study of Painted Works of Art
AU - Rohani, Neda
AU - Pouyet, Emeline
AU - Walton, Marc Sebastian
AU - Cossairt, Oliver Strides
AU - Katsaggelos, Aggelos K
N1 - Funding Information:
This collaborative initiative is part of NU-ACCESS≫s broad portfolio of activities, made possible by generous support of the Andrew W. Mellon Foundation. The authors would like to thank Jessica Chloros and Valentine Talland Associate Objects Conservator at the Isabella Gardner Museum (Boston, USA) for motivating non-invasive analyses on the illuminated manuscript.
Publisher Copyright:
© 2018 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
PY - 2018/8/20
Y1 - 2018/8/20
N2 - Nonlinear unmixing of hyperspectral reflectance data is one of the key problems in quantitative imaging of painted works of art. The approach presented is to interrogate a hyperspectral image cube by first decomposing it into a set of reflectance curves representing pure basis pigments and second to estimate the scattering and absorption coefficients of each pigment in a given pixel to produce estimates of the component fractions. This two-step algorithm uses a deep neural network to qualitatively identify the constituent pigments in any unknown spectrum and, based on the pigment(s) present and Kubelka–Munk theory to estimate the pigment concentration on a per-pixel basis. Using hyperspectral data acquired on a set of mock-up paintings and a well-characterized illuminated folio from the 15th century, the performance of the proposed algorithm is demonstrated for pigment recognition and quantitative estimation of concentration.
AB - Nonlinear unmixing of hyperspectral reflectance data is one of the key problems in quantitative imaging of painted works of art. The approach presented is to interrogate a hyperspectral image cube by first decomposing it into a set of reflectance curves representing pure basis pigments and second to estimate the scattering and absorption coefficients of each pigment in a given pixel to produce estimates of the component fractions. This two-step algorithm uses a deep neural network to qualitatively identify the constituent pigments in any unknown spectrum and, based on the pigment(s) present and Kubelka–Munk theory to estimate the pigment concentration on a per-pixel basis. Using hyperspectral data acquired on a set of mock-up paintings and a well-characterized illuminated folio from the 15th century, the performance of the proposed algorithm is demonstrated for pigment recognition and quantitative estimation of concentration.
KW - deep neural network classification
KW - heritage science
KW - nonlinear unmixing Kubelka–Munk theory
KW - visible hyperspectral imaging
UR - http://www.scopus.com/inward/record.url?scp=85050608277&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=85050608277&partnerID=8YFLogxK
U2 - 10.1002/anie.201805135
DO - 10.1002/anie.201805135
M3 - Article
C2 - 29940088
AN - SCOPUS:85050608277
VL - 57
SP - 10910
EP - 10914
JO - Angewandte Chemie - International Edition
JF - Angewandte Chemie - International Edition
SN - 1433-7851
IS - 34
ER -