wpotrzebowski
commented
2 years ago SVD is also used in Bio-SANS to get initial idea of number of components (oligomers) in 1D scattering curve. I've also worked with MCR method (Identification of Intermediates States with Multivariate Curve Resolution section in https://pubs.acs.org/doi/10.1021/acsnano.0c03569) , which has some advantages over PCA/SVD. At the time my collegue implemted in R but I think there is also a python implementation of the algorithm.
Principal component analysis (PCA) gives a handle to calculate the direction and degree of anisotropy in a 2D scattering image. It can be performed based on a Singular Value Decomposition (SVD) of the data. The eigenvector belonging to the largest eigenvalue gives the preferred scattering direction and the eigenvalue scales with the maximal variance. This tutorial gives further information on the relation between PCA and SVD: https://arxiv.org/pdf/1404.1100.pdf