% IMPORTANT: The following is UTF-8 encoded.  This means that in the presence
% of non-ASCII characters, it will not work with BibTeX 0.99 or older.
% Instead, you should use an up-to-date BibTeX implementation like “bibtex8” or
% “biber”.

@INPROCEEDINGS{Lhrs:151502,
      author       = {Lührs, Anna},
      title        = {{H}ybrid parallelization of a seeded region growing
                      segmentation of brain images for a {GPU} cluster},
      address      = {Berlin},
      publisher    = {VDE Verlag},
      reportid     = {FZJ-2014-01440},
      isbn         = {978-3-8007-3579-2},
      pages        = {8},
      year         = {2014},
      comment      = {ARCS 2014: 27th International Conference on Architecture of
                      Computing Systems - Workshop Proceedings},
      booktitle     = {ARCS 2014: 27th International
                       Conference on Architecture of Computing
                       Systems - Workshop Proceedings},
      abstract     = {The introduction of novel imaging technologies always
                      carries new challenges regarding the processing of the
                      captured images. Polarized Light Imaging (PLI) is such a new
                      technique. It enables the mapping of single nerve fibers in
                      postmortem human brains in unprecedented detail. Due to the
                      very high resolution at sub-millimeter scale, an immense
                      amount of image data has to be reconstructed
                      three-dimensionally before it can be analyzed. Some of the
                      steps in the reconstruction pipeline require a previous
                      segmentation of the large images. This task of image
                      processing creates black-and-white masks indicating the
                      object and background pixels of the original images. It has
                      turned out that a seeded region growing approach achieves
                      segmentation masks of the desired quality. To be able to
                      process the immense number of images acquired with PLI, the
                      region growing has to be parallelized for a supercomputer.
                      However, the choice of the seeds has to be automated in
                      order to enable a parallel execution. A hybrid
                      parallelization has been applied to the automated seeded
                      region growing to exploit the architecture of a GPU cluster.
                      The hybridity consists of an MPI parallelization and the
                      execution of some well-chosen, data-parallel subtasks on
                      GPUs. This approach achieves a linear speedup behavior so
                      that the runtime can be reduced to a reasonable amount.},
      month         = {Feb},
      date          = {2014-02-25},
      organization  = {27th International Conference on
                       Architecture of Computing Systems,
                       Lübeck (Germany), 25 Feb 2014 - 28 Feb
                       2014},
      cin          = {JSC / JARA-HPC},
      cid          = {I:(DE-Juel1)JSC-20090406 / $I:(DE-82)080012_20140620$},
      pnm          = {411 - Computational Science and Mathematical Methods
                      (POF2-411) / SMHB - Supercomputing and Modelling for the
                      Human Brain (HGF-SMHB-2013-2017) / SLNS - SimLab
                      Neuroscience (Helmholtz-SLNS)},
      pid          = {G:(DE-HGF)POF2-411 / G:(DE-Juel1)HGF-SMHB-2013-2017 /
                      G:(DE-Juel1)Helmholtz-SLNS},
      typ          = {PUB:(DE-HGF)8 / PUB:(DE-HGF)7},
      url          = {https://juser.fz-juelich.de/record/151502},
}