Multi-channel boosting and multi-scale localization-based tracking of dense malarial sporozoites
Astha Jaiswal, Konrad Beyer, Friedrich Frischknecht, Karl Rohr · 2018
Tracking of biological objects at microscopic scale is a powerful tool for analyzing the dynamics of biological systems. However, particularly a high density of highly motile objects make tracking a challenging task. In this work, we propose a tracking approach which utilizes multiple frames, multiple channels, and multiple scales. Our approach employs a multi-channel boosting scheme to cope with high density of objects. Clustering and unclustering of objects is handled using many-to-one and one-to-many associations as well as identity resolution. Multi-scale information is used to estimate the position of clustered objects. We have successfully applied our approach for tracking of malarial sporozoites in real microscopy image sequences and performed a quantitative comparison with state-of-the-art approaches.