Metrology-Based Statistical Framework for Monitoring Changes in Hyperspectral Datasets of Artworks

Abstract

Hyperspectral imaging is a valuable technique for preserving and monitoring cultural heritage objects. Over time, artworks can undergo subtle changes due to environmental factors or handling during loan periods, making it critical to detect these changes for effective conservation. However, maintaining identical acquisition conditions across time—due to variations in spectral and spatial resolution, illumination, and calibration—is often impractical, making pixel-wise comparison unfeasible. To address this, we introduce a metrology-based statistical framework that allows detecting changes or anomalies in hyperspectral datasets of the same object acquired at different time intervals under different acquisition condition. The nonlinear framework uses Kullback-Leibler pseudo-divergence and Mahalanobis distance to enable robust comparison and detection of significant changes in hyperspectral data, providing a valuable tool for effectively monitoring and preserving cultural heritage artifacts.