Emerging Field 3 “Model-based Imaging”

Motivated by applications in magnetic resonance imaging (related to fingerprinting and multimodality), electron and X-ray microscopy, as well as the identification of particle properties from macroscopic data, the research in EF3 has its focus on the following main directions: (i) Integrated physics-based imaging (IPI) including model learning based reconstruction; (ii) Information extraction from multimodal data; (iii) Advanced variational methods and solver design for problems with huge-scale data.

 

One of the main perspective goals of EF 3 is to extend the notion of integrated physics-based imaging to other modalities such as X-ray microscopy, where the state system will be related to Maxwell’s equations, or electron-holography, where phase-information, leading to a notoriously complicated retrieval problem, becomes relevant. Another major challenge which will come into the focus of EF 3 is the quantification of uncertainties for clustering problems and of geometric information in imaging, such as segmentations or other topological structures.

 

In general, success in these themes requires major advances in modeling, analysis, hierarchical or non-smooth optimization, statistics-based learning and high-performance computations. The long-term vision of EF 3 is to establish an internationally recognized Berlin-wide platform for quantitative image processing and imaging, where innovation in the applied sciences and device engineering are driven by new mathematical developments.

 

Projects marked with an asterisk (*) are short-term one-year pilot projects.

 

Scientists in Charge: Michael Hintermüller, Vladimir Spokoiny, Gabriele Steidl

The projects EF3-7 and EF3-12 have been moved to Application Area 5: Variational Problems in Data-Driven Applications.

 

Successfully completed projects: