Tracking virus particles in fluorescence microscopy images via a particle Kalman filter

Tracking fluorescent particles in microscopy image sequences is pivotal in obtaining quantitative characterizations of the dynamical processes underlying these fluorescent structures. We have developed a probabilistic tracking approach that combines the Kalman filter with principles of the particle filter. To generate samples, we use an elliptical approximation of a Gaussian density. Each sample is weighted according to an image likelihood and the image support. The performance of our tracking approach has been evaluated using multi-dimensional synthetic as well as real microscopy image data. The approach yields a more accurate performance at very competitive computation times compared to previous probabilistic approaches.