A Study on Tensor and Matrix Models for Super-Resolution Fluorescence Microscopy

José Henrique de Morais Goulart, Laure Blanc-Féraud, E. Debreuve, Sebastien Schaub

Abstract: Super-resolution techniques for fluorescence microscopy are invaluable tools for studying phenomena that take place at sub-cellular scales, thanks to their capability of overcoming light diffraction. Yet, achieving sufficient temporal resolution for imaging live-cell processes remains a challenging problem. Exploiting the temporal fluctuations (blinking) of fluo-rophores is a promising approach that allows employing standard equipment and harmless excitation levels. In this work, we study a novel constrained tensor modeling approach that takes this temporal diversity into account to estimate the spatial distribution of fluorophores and their overall intensities. We compare this approach with an also novel matrix-based formulation which promotes structured sparsity via a continuous approximation of the cardinality function, as well as with other state-of-the-art methods.