Multivariate analysis of hyperspectral hard X‐ray images

This article describes methods to analyse and process hyperspectral hard X-ray imaging data. We focus on the use of multivariate techniques that exploit the spectral information to make informed decisions on the material content within each pixel of an X-ray image. These analysis methods have the ability to auto-segment data without prior knowledge of the sample composition or structure, and are particularly useful for studying completely unknown, diluted or complex specimens. We demonstrate the methods on a variety of hard X-ray images including X-ray fluorescence and absorption data recorded using a hard X-ray imaging spectrometer. The multivariate methods described are very powerful with the ability to segment, distinguish and, in some cases, identify different materials within a single X-ray image. Potential uses of hyperspectral X-ray imaging are discussed varying from materials science to industrial or security applications. Copyright (c) 2013 John Wiley & Sons, Ltd.