Improving active contours for segmentation and tracking of motile cells in videomicroscopy

We describe a method to segment and track motile biological cells in videomicroscopy data, based on parametric active contours. We propose improvements designed to overcome obstacles in applications to cellular imaging. To allow detection of low contrast pseudopodia, we use an edge map based on the local intensity deviation that better reflects cell/background boundaries. To allow correct segmentations of cells in contact, we introduce a repulsion effect between contours. The method performance is illustrated on real data.