Working Group Algorithms for Computer Vision, Imaging and Data Analysis

Andreas Weinmann

Andreas Weinmann received the Diplom degree in Mathematics and Computer Science (minor) from Technische Universität München, in 2006, and the PhD degree from Technische Universität Graz, in 2010. He worked as a researcher with the Technische Universtät München and with the Helmholtz Center Munich. Since 2015, he is a professor for Mathematics with focus on Image Processing at Hochschule Darmstadt.

His research interests lie in the intersection of Mathematics, Computer Science and Engineering. In particular, he is interested in the development and application of algorithms in Signal and Image Processing/Computer Vision and in Biomedical Imaging. Considered application domains are for instance Magnetic Particle Imaging and Raman Spectroscopy. Further interests are the modeling and analysis of the developed algorithms. The employed methodology includes numerical optimization, variational regularization of inverse problems and machine learning based on Neural Networks.

Website | Andreas Weinmann

Thomas März

Thomas März received his Diplom degree in Applied Mathematics with minors Computer Science and Electrical Engineering in 2005, and his PhD degree in 2010 both from Technische Universität München. He worked as postdoctoral researcher at the Oxford Centre for Collaborative Applied Mathematics from 2010 to 2014. After that, from 2014 to 2018, he was as senior software engineer with Roxar Ltd. (Oxford) and worked on the further development of Roxar's oil reservoir simulator. Since 2018 he is professor of Mathematics at Hochschule Darmstadt with focus on Applied Mathematics in Engineering.

His research interests are in the overlap of the areas of Numerical Analysis, Mathematical Modeling, Computer Science and Engineering. He focuses on the mathematical modeling of technical problems and their computer aided solution through the development of robust and efficient numerical algorithms. Applications include interpolation of missing data in signals and images (image inpainting), denoising on curved surfaces as well as inverse problems in imaging.

Website | Thomas März

Vladyslav Gapyak

Vladyslav Gapyak received both his Bachelor's and Master's degree in Mathematics at the University of Padua in 2019 and in 2021. Since 2022 he is employed at Hochschule Darmstadt.

His current work deals specifically with the problem of modeling and regularizing the MPI imaging modality.
He considers classical regularization techniques as well as recently developed approaches based on machine learning.

Vladyslav Gapyak

Georg Frey

Georg Frey received a Bachelor's in Applied Computer Science from Duale Hochschule Baden-Württemberg in 2019, and a Master's degree in Data Science from Hochschule Darmstadt in 2021. Since 2022 he is employed at Hochschule Darmstadt working on the analysis of Raman spectroscopic data.

He considers classical as well as modern machine-learning methods in a dual resolution setup.

Georg Frey

Tamara Dieter

Tamara Dieter received a bachelor's degree in mathematics in 2018 and a master's degree in mathematics in 2020, both from the University of Applied Sciences Darmstadt. Since 2021, she is employed in a collaborative project between the University and the German Aerospace Center.

She focuses on visual object detection, specifically on drone detection in the context of perimeter protection, using machine learning methods (such as Deep Neural Networks) and contextual information.

Tamara Dieter


Topics

We all acquire data to get information, e.g., we take pictures with our mobile phone's camera. Physicians use imaging devices to infer on person's health conditions, scientists use microscopic data to learn about small organisms, and engineers use imaging based on spectroscopic methods to do diagnostics for chemical reactions.

What is common to all these situations is that the data measured does not directly give the information one is interested in; the measured data needs to be processed and the required information needs to be extracted.

The central topic of ACIDA is the development and application of algorithmic solutions to such kind of problems in Computer Vision, Imaging and related Data Analysis problems.

Considered research topics are as follows:

  • Variational methods of Bayesian MAP estimation type for segmentation, denoising, regularization and reconstruction
  • PDE and/or variational methods for data completion/inpainting problems; PDE-models on curved surfaces
  • Reconstruction techniques for the emerging imaging modality Magnetic Particle Imaging (MPI)
  • Analysis of Raman spectroscopic data using taylored classical and machine-learning based approaches
  • Algorithms for nonlinear data such as phase angles, directions, poses, and positive matrices (nerve fibre orientations) based on differential geometry;
  • Developmemt of algorithmic solutions for Computer Vision problems based on deep learning frameworks (e.g., involving specific segmentation/object tasks).

From a methodological side, we use methods of scientific computing, in particular of numerical optimization. We develop and employ iterative schemes, dynamic programming as well as derivative based methods.


Publications