Pearson correlation-based method on hyperspectral images for the study of similarity of pigments and dyes

Authors

Cou, Corentin; de La Codre, Hortense; Granier, Xavier; Mounier, Aurélie

Contributors

Laboratoire Photonique, Numérique et Nanosciences (LP2N); Université de Bordeaux (UB)-Institut d'Optique Graduate School (IOGS)-Centre National de la Recherche Scientifique (CNRS); L'information visuelle et textuelle en histoire de l'art : nouveaux terrains, corpus, outils (InVisu); INHA-Centre National de la Recherche Scientifique (CNRS); Melting the frontiers between Light, Shape and Matter (MANAO); Laboratoire Bordelais de Recherche en Informatique (LaBRI); Université de Bordeaux (UB)-École Nationale Supérieure d'Électronique, Informatique et Radiocommunications de Bordeaux (ENSEIRB)-Centre National de la Recherche Scientifique (CNRS)-Université de Bordeaux (UB)-École Nationale Supérieure d'Électronique, Informatique et Radiocommunications de Bordeaux (ENSEIRB)-Centre National de la Recherche Scientifique (CNRS)-Inria Bordeaux - Sud-Ouest; Institut National de Recherche en Informatique et en Automatique (Inria)-Institut National de Recherche en Informatique et en Automatique (Inria)-Laboratoire Photonique, Numérique et Nanosciences (LP2N); Université de Bordeaux (UB)-Institut d'Optique Graduate School (IOGS)-Centre National de la Recherche Scientifique (CNRS)-Institut d'Optique Graduate School (IOGS)-Centre National de la Recherche Scientifique (CNRS); Archéosciences Bordeaux; Université de Bordeaux (UB)-Université Bordeaux Montaigne (UBM)-Centre National de la Recherche Scientifique (CNRS); Institut des Sciences Moléculaires (ISM); Université Montesquieu - Bordeaux 4-Université Sciences et Technologies - Bordeaux 1 (UB)-École Nationale Supérieure de Chimie et de Physique de Bordeaux (ENSCPB)-Institut de Chimie - CNRS Chimie (INC-CNRS)-Centre National de la Recherche Scientifique (CNRS); 80|Prime Smart 3D + 13

Published in

International conference on analytical techniques in art and cultural heritag... >

Publisher

HAL CCSD

Abstract

International audience ; The emergence of hyperspectral cameras (NIR-VIS) has made it possible to acquire millions of spectra on samples. This has generated a need to use data processing and visualization methods because manual observation is no longer possible. However, when the data becomes complex with variations in recipes, intensity or mixture within the same dye or pigment, common methods of segmentation no longer work very well (classification according to the intensity and not the shifts visible on the spectra for example). Pearson correlation-based data treatment is developed and discussed in this paper. We find the use of reflectance spectra to answer questions in many cases. For example, the study of 18th century Aubusson tapestries dyes by crossing hyperspectral imaging and other non-invasive analyses methods is carried out for dye identification purposes [1]. Another illustration is the use of hyperspectral imaging on Iznik ceramics tiles inside the Saint-Maurice Residence (Cairo). Patterns similar to those of the residence exist dotted around Cairo [2]. Their study allows for traceability in the context of reuse. But in both cases, problems arise due to huge amounts of data for variations of the same pigments and dyes [3], and therefore we needed to develop a new method to reduce them. This study proposes a method to enhance the robustness of hyperspectral images processing and reduce the amount of data by generating tools for a similarity study between studied spectra and a database. The first step is the creation of a database of key spectra used for correlations. Then, some pre-processing are applied to the studied hyperspectral imaging (like spatial filtering for denoising). The main point of our method is that we compute a Pearson correlation coefficient between the studied spectra and each of the key spectra from the database. These new values obtained can be used for common methods of segmentation and visualisation. Our processes have been applied to different cases. After testing it on ...