2. Data diagnostics in imaging

Spokesperson: Peter Fratzl (MPI of Colloids and Interfaces)

Co-Spokesperson: Peter Benner (MPI for Dynamics of Complex Technical Systems)

New imaging methods based on light or x-ray spectroscopy (or scattering) provide important information on structure and properties of (new) materials. Such approaches are particularly important for inhomogeneous materials close to real-life conditions. The associated challenge is that, on the one hand, collecting such data is very expensive (for example, using precious time on a synchrotron source) but, on the other, the size of relevant data sets becomes so huge that their quality cannot be easily assessed during the measurement. Therefore, an effort has been started to develop tools for online diagnostics that enables a rapid assessment of data quality without a full blown analysis of the data.

The principle of SAXS tomography is to determine complete (3D) SAXS patterns for every pixel (or voxel) of the scanned plane (or tomogram). This result is an up to six-dimensional information (three dimensions from tomography and three dimensions for the SAXS patterns in each voxel). Typically, the number of SAXS frames to be taken for one tomographic reconstruction is in the order of more than one million. Considering that one SAXS frame has a size of several megabytes, the amount of data for one tomographic reconstruction is in the order of terabytes.

We aim to reduce the SAXS data already during the measurements applying algorithms on scalars resulting from a linear superposition (similar to normal tomography, where the logarithm of the transmission follows a linear superposition principle).

Here are the individual projects and respective members:

2.1. Data diagnostics in SAXS tomography - Peter Fratzl, Wolfgang Wagermaier (MPI of Colloids and Interfaces), Peter Benner (MPI for Dynamics of Complex Systems), Martin Stoll (TU Chemnitz)

2.2. Automated crystallographic analysis of atom probe detector maps - Baptiste Gault (MPI for Iron Research), Tristan Bereau (MPI for Polymer Research)

2.3. Big data enabled true analytical atomic scale tomography - Leigh Stephenson (MPI for Iron Research), Markus Rampp (MPI for Data Facility), Stefan Bauer (MPI for Intelligent Systems)

2.4. Physical inference from multidimensional photoemission data - Ralph Ernstorfer, Patrick Xian (Fritz Haber Institute), Stefan Bauer, Vincent Stimper (MPI for Intelligent Systems )

2.5. (NEW) 3D shape-based object segmentation of large tomographic datasets - Luca Bertinetti, Richard Weinkamer, Wolfgang Wagermaier (MPI for colloids and Interfaces ), Markus Kühbach, (MPI for Iron Research), Markus Rampp, Andreas Marek, Nicolas Fabas (MP Computing and Data Facility)