Cell tracking using particle filters and level sets

We propose an algorithm to track moving cells and microbes in a video. A major challenge in tracking living cells is that their movement is often nonlinear which causes problems in case of approaches using the generic particle filter (GPF) framework. In order to overcome this problem, we propose the use of an auxiliary particle filtering (APF) algorithm with dynamic variance adaptation of the posterior distribution to account for nonlinear movements. The object of interest in each frame is segmented using level sets. The proposed tracking algorithm is tested on real data and the tracking performance is compared with that of GPF and APF without dynamic variance adaptation. Experimental results show that the proposed algorithm tracks more accurately compared to GPF and APF without variance adaption, with lesser number of particles, thereby reducing the running time.