Fig. 4From: Optimal principal component analysis of STEM XEDS spectrum imagesStandard characterization of principal component decomposition—scree plots of the components obtained by a, b unweighted decomposition, c, d weighted decomposition and e, f weighted decomposition with the filtering pre-treatment. a, c, e show the variances of synthetic noise-free (TRUE) datasets, while (b, d, f) show the variances for the noisy synthetic (NOISE) and experimental (EXP) overlapped with the TRUE ones. Note that the variances of components are plotted in the logarithmic scaleBack to article page