Fig. 1

Schematic illustration of the registration algorithm. Top, from left: The algorithm takes the reconstructed STORM data (in the form of fluorophore x, y localizations) and applies a rigid transformation to the fluorophore localizations to perform registration between channels, producing localizations corrected for drift. Bottom, from left: Initially, fiducial markers are identified from each channel using two detection criteria: (1) The mean number of localizations present within a local neighborhood over time exceeds a given threshold, and (2) The variance of the number of localizations within the neighborhood over time is below a given threshold. As a result, non-blinking constantly emitting sources corresponding to the fiducial markers are retained. The rigid transformation between the channels is then estimated to align the fiducial markers, using the Iterative Closest Point algorithm [32,33,34]. The ICP algorithm is used to align two-point sets by minimizing the distance between corresponding points. It iteratively updates the transformation between the two sets until convergence using point-to-point correspondences and estimates a rigid transformation combining rotation and translation through a root mean square distance minimization