GT ADG-Systèmes DynamiquesStochastic optical reconstruction microscopy
Jean-François Rupprecht (UPMC, National University of Singapore)
Tuesday 03 November 2015 14:00 - Orléans - Salle de Séminaire
I will present a probabilistic model that predicts the temporal resolution limit in stochastic imaging techniques. The recently developed stochastic optical reconstruction microscopy allows to reach a spatial resolution below the nanometer diffraction limit (and for its inventors, it also allowed to obtain the 2014 Nobel price in Chemistry). This imaging technique relies on the superposition of large stack of snapshots, that each contains a scarce number of super-resolved light emission events. As the subset of observed pixels is random, the minimal number of snapshots required to obtain a reliable recollected image (called temporal resolution limit) is a random variable. First, if one observation per pixel is sufficient (as in a PALM experiment), we show that the temporal resolution follows a non-linear N log(N) behavior with the number of pixels N. This behavior corresponds to the "coupon collector" result. Secondly, we focus on a new holographic technique to characterize electromagnetic fields using Brownian nanoparticles as local probes. Due to noise in the holographic localization procedure, we need to accumulate a large number of observations per pixel. In this case, the temporal resolution is shown to be proportional to the number of pixels N. We point out that the determination of the electromagnetic field profile is a non-parametric estimation problem.