Hierarchical clustering of spectral images with spatial constraints for the rapid processing1of large and heterogeneous datasets from ancient material studies
Abstract
The study of very complex and heterogeneous materials, such as those encountered in the science of ancient materials, benefits from the wealth of information provided by the acquisition and exploitation of full spectrum images, i.e. spectral images. In order to obtain a high dynamic range in both the spatial dimensions and composition, great efforts have made it possible to considerably accelerate data collection and increase the average size of a single data set, each image reaching up to several tens of GB. Rapid processing is now required to allow feedback during data collection, within the short time available for instruments and samples. Here we propose an approach combining hierarchical clustering and spatial constraint. Spatial constraints allow both a significant reduction in the computational cost of segmentation and a certain level of robustness with respect to the signal-to-noise ratio: the prior knowledge injected by the spatial constraint partially compensates for the increase in noise level; hierarchical clustering provides a statistically sound and known framework that allows accurate reporting of the instrument noise model. We illustrate the proposed algorithm on a X-ray fluorescence spectral image collected on an ca. 100 Myr fossil fish, as well as on simulated data to assess the sensitivity of the results to the noise level. It can be foreseen how such an approach could simultaneously lead to an increase in the spatial definition of the collected spectral image and to a reduction in the potentially harmful radiation dose density to which the samples are subjected.
Origin | Files produced by the author(s) |
---|