Skip to Main Content
Tracking of multiple objects in biological image data is a challenging problem due largely to poor imaging conditions and complicated motion scenarios. Existing tracking algorithms for this purpose often do not provide sufficient robustness and/or are computationally expensive. In this paper we propose a new object detection scheme, based on importance sampling from image intensity distributions, and show how it can be easily incorporated into a probabilistic tracking framework based on Kalman or particle filtering. Experiments on synthetic as well as real fluorescence microscopy image data from different biological studies show that the resulting tracking algorithm yields smaller localization errors at much lower execution times compared to other available methods.