Improving 3d tracking in microscopy by joint estimation of kinetic and image models

Abstract—We present a new way to combine 3D microscopy image models with kinetic models of targets in order to build an efficient multiple particle tracking algorithm. The method is dedicated to the task of tracking small particles in a dense biological environment, in which they can appear fused. We propose to enrich a Bayesian kinetic tracking score with an additional feature related to the image model. The feature is a statistical function based on both the noise distribution of acquisition systems and adaptive target profiles. The profiles estimation takes into account the Point Spread Function of the microscope. A systematic performance analysis and a behavior study are given on synthetic realistic data set. A real 3D case is demonstrated and the benefits of the approach for the tracking task in these images are illustrated. I