Location: MIC-Arena, Maria-Goeppert-Str. 3, 23562 Lübeck
Title: "Variational Image Processing for Electron Microscopy"
This talk introduces several recent methods from variational image
processing tailored to data from electron microscopy. In particular,
specialized methods for variants of three fundamental image processing
problems will be discussed: denoising, segmentation and registration.
First, we demonstrate that non-rigid registration techniques can be used
to achieve sub-picometre precision measurements of atom positions from a
series of scanning transmission electron microscopy (STEM) images at
atomic scale. Then, we show how non-local denoising methods like the
Block-matching and 3D filtering algorithm (BM3D) can be extended to
exploit the periodicity of crystalline structures visible in atomic
scale images. This needs knowledge about the underlying crystal lattice,
for which we introduce a real space method for unsupervised primitive
unit cell extraction. Moreover, to cover cases of micrographs that show
materials with discontinues in the periodicity (like in case of multiple
grains), we developed a novel method for multi-phase segmentation of
images based on high-dimensional local feature vectors. The denoising
and segmentation techniques can be generalized to hyperspectral data,
like EELS or EDX.
Finally, we propose a novel method that jointly estimates the direction
dependent STEM distortions in each image of a STEM image series at
atomic scale and reconstructs the underlying atomic grid of the material
by fitting the atoms with bump functions