Multiple Hypothesis Tracking for Cluttered Biological Image Sequences

摘要: In this paper, we present a method for simultaneously tracking thousands of targets in biological image sequences, which is major importance modern biology. The complexity and inherent randomness the problem lead us to propose unified probabilistic framework particles microscope images. includes realistic models particle motion existence fluorescence features. For track extraction process per se, very cluttered conditions motivate adoption multiframe approach that enforces decision robustness poor imaging random target movements. We tackle large-scale nature by adapting multiple hypothesis algorithm proposed framework, resulting with favorable tradeoff between model computational cost procedure. When compared state-of-the-art techniques bioimaging, shown be only providing high-quality results despite critically dense presence. thus demonstrate benefits advanced Bayesian accurate modeling dynamical processes, promising further developments domain.