Fig. 8From: Optimal principal component analysis of STEM XEDS spectrum imagesElemental maps extracted from (a) raw and (b) filtered noisy synthetic datasets; c–e represent elemental maps extracted from the reconstructed datasets according to the (c) unweighted PCA, (d) weighted PCA and (e) filtered and weighted PCA treatments. The number of principal components used in reconstruction was 22 for (c), 9 for (d) and 7 for (e). For fair comparison, all maps are displayed with identical contrast settings. The PCA treatment improves noticeably the visibility of the chemical components, especially the weak-signal elements—Ta and Hf. Note that the treatment variants (c) and (d) cause the artificial intermixing of the Ta and Hf layersBack to article page