Fig. 1.
(a) Workflow for the conventional denoising, KK, PEC, and SEC, where m is the total number of spatial pixels (“flattened") and n is the number of frequency channels. (b) Workflow for the fKK-EC where the processing steps are performed on basis vectors rather than the underlying spectra. (c) Workflow for the ML:fKK-EC in which only the training data is processed via the fKK-EC and regression is used to transform new data.