Figure 4.

Clustering and Postprocessing. a) Induced clustering, where each pixel is assigned to the row of S (see Figure 3) where it has the highest coefficient. b) Visualisation of the signal refinement procedure in postprocessing. The movie matrix is projected onto the top-two signals, t⁽⁰⁾, t⁽¹⁾ ∈ T. Every black dot corresponds to a time series (pixel), and all time series that have been selected into T are shown as red dots. After averaging over the time series in the coloured areas, signals are shifted from t⁽⁰⁾ to the mean ${\hat{t}}^{\left(0\right)}$ and from t⁽¹⁾ to the mean ${\hat{t}}^{\left(1\right)}$c) Glomerular map: Induced clustering based on $Ŝ$, i.e. after postprocessing. Pixels that do not participate in any mean signal are set to white.