Combinatorial Image Analysis
Research Goal and Method
Our research is in the area of image analysis, with methods reaching into the fields of machine learning and optimization, and with applications in computer vision and bio-medical imaging.
The goal of our research is to define and implement algorithms for image analysis that surpass the human in terms of accuracy and processing time. This research goal is motivated by the broad impact such algorithms can have on the society, for instance, by accelerating scientific research in biology and medicine through intelligent microscopy, and by enabling the design of autonomous devices in engineering.
In pursuit of this goal, we define, advance and study mathematical abstractions of image analysis tasks in the form of optimization problems. We define, implement and apply algorithms for solving these problems exactly or approximately, and we compare the accuracy and processing time of these algorithms to that of humans in terms of existing benchmarks. This approach connects our research to a rich body of knowledge in the fields of combinatorial optimization and convex optimization on which we build.
The Minimum Cost Multicut Problem in the field of image analysis
Minimum Cost Multicuts for Image Segmentation
Keuper et al. 2015
Andres et al. 2012
Andres et al. 2011
Minimum Cost Multicuts for Multi-Target Tracking
Jug, Levinkov et al. 2015
Tang et al. 2015
Minimum Cost Multicuts for Human Body Pose Estimation
Insafutdinov et al. 2016
Pishchulin et al. 2016
Lifting of Multicuts
Andres, Fuksová and Lange 2016