I have moved to the Computer Science and Artificial Intelligence Laboratory at the Massachusetts Institute of Technology (MIT-CSAIL). This is my new website.

Jan Stühmer received his Diploma degree (with distinction) in Computer Science from Dresden University of Technology in 2010. In his diploma thesis he developed a novel method that allows dense 3D reconstructions with a handheld camera. In 2014/2015 he stayed as a research intern with Microsoft Research Cambridge and in 2013 as a visiting student researcher at the Applied Geometry Lab at Caltech. From 2005 to 2009 he was with the group of Carl-Philipp Heisenberg at the Max Planck Institute of Molecular Cell Biology and Genetics. Since October 2010 he is a Ph.D. student in the Research Group for Computer Vision, Image Processing and Pattern Recognition headed by Prof. Daniel Cremers. Jan is coadvised by Prof. Peter Schröder (Caltech) and supported by the Institute for Advanced Study.

Geometry Processing, Segmentation, Convex Optimization, Variational Methods, Biomedical Image Processing, Probabilistic Models, GPU Programming

I am working in the following research areas:

We propose a novel tracking method that directly performs model based tracking on the raw infrared signal of a Time-Of-Flight camera which allows us to reconstruct the object’s depth at an order of magnitude higher frame-rate. Even when the depth reconstruction fails due to fast motion of the object, our method can track the moving object.

Especially in biomedical image segmentation and denoising the structures of interest show a thin and fine detailed shape. While state-of-the-art segmentation methods perform well for segmenting compact objects, their performance on thin structures is often not satisfying. The commonly used length regularizer suppresses small structures and the correct topology cannot be reconstructed. To overcome this limitation, we introduce a novel algorithmic framework, that allows to preserve the connectivity of the object. We show that our method can be successfully applied to medical image segmentation problems in angiography and retinal blood vessel extraction, where thin structures otherwise would not be preserved by boundary length regularizers.

We present a novel variational approach to estimate dense depth maps from multiple images in real-time using a hand-held camera. Robust penalizers for both data term and regularizer allow to preserve discontinuities in the depth map. We demonstrate that the integration of multiple images substantially increases the robustness of estimated depth maps to noise in the input images.

In this project, we quantified the role of cell-cell adhesion for collective migration during embryogenesis of the developing zebrafish. To allow an accurate quantification of cell migration in vivo, we developed a segmentation and tracking framework which is very robust to the salt and pepper noise observed in confocal laser microscope image data.