Fig. 12From: Optimal principal component analysis of STEM XEDS spectrum imagesScatter plots for the experimental (a, d, g, j), synthetic noisy (b, e, h, k) and noise-free (c, f, i, l) datasets described in "Multi-component object for spectrum-imaging" section. All datasets were subjected to the filtering pre-treatment followed by the weighted principal components decomposition as described in "PCA with smoothing filter pre-treatment" section. The scatter plots are composed of the 1st and 2nd (a–c); 2nd and 3rd (d–f), 3rd and 4th (g–i) and 4th and 5th (j–l) principal components. The spectral positions of the pure latent factors whenever they can be identified in the scatter plots are marked by the red pointsBack to article page