@InProceedings{	  gustavsson.ea:09a,
  author	= {David Gustavsson and Kim Steenstrup Pedersen and Mads
		  Nielsen},
  title		= {A SVD Based Image Complexity Measure},
  booktitle	= {International Conference on Computer Vision Theory and
		  Applications (VISAPP'09)},
  year		= {2009},
  project	= {VISIONTRAIN,NISA}
}

@InProceedings{	  gustavsson.ea:09b,
  author	= {David Gustavsson and Kim Steenstrup Pedersen and Francois
		  Lauze and Mads Nielsen},
  title		= {On the rate of structural change in scale spaces },
  booktitle	= {Proceedings of Scale Space and Variational Methods in
		  Computer Vision {SSVM}},
  year		= {2009},
  project	= {VISIONTRAIN,NISA}
}

@Article{	  cuzol.pedersen.ea:08,
  author	= {Anne Cuzol and Kim Steenstrup Pedersen and Mads Nielsen},
  title		= {Field of Particle Filters for Image Inpainting},
  journal	= {Journal of Mathematical Imaging and Vision},
  year		= {2008},
  volume	= {31},
  number	= {2-3},
  pages		= {147--156},
  project	= {VISIONTRAIN,NISA},
  note		= {\url{http://dx.doi.org/10.1007/s10851-008-0072-7}}
}

@InProceedings{	  gustavsson.fundana.overgaard.nielsen:08b,
  author	= {David Gustavsson and Ketut Fundana and Niels Chr.
		  Overgaard and Anders Heyden and Mads Nielsen},
  title		= {Nonrigid Object Segmentation and Occlusion Detection in
		  Image Sequences},
  booktitle	= {3rd International Conference on Computer Vision Theory and
		  Applications ({VISAPP} 08)},
  year		= {2008},
  abstrac	= {We address the problem of nonrigid object segmentation in
		  image sequences in the presence of occlusions. The proposed
		  variational segmentation method is based on a region-based
		  active contour of the Chan-Vese model augmented with a
		  frame-to-frame interaction term as a shape prior. The
		  interaction term is constructed to be pose-invariant by
		  minimizing over a group of transformations and to allow
		  moderate deformation in the shape of the contour. The
		  segmentation method is then coupled with a novel
		  variational contour matching formulation between two
		  consecutive contours which gives a mapping of the
		  intensities from the interior of the previous contour to
		  the next. With this information occlusions can be detected
		  and located using deviations from predicted intensities and
		  the missing intensities in the occluded regions can be
		  reconstructed. After reconstructing the occluded regions in
		  the novel image, the segmentation can then be improved.
		  Experimental results on synthetic and real image sequences
		  are shown.},
  project	= {VISIONTRAIN}
}

@InProceedings{	  gustavsson.ea:08a,
  author	= {David Gustavsson and Kim Steenstrup Pedersen and Mads
		  Nielsen},
  title		= {Multi-Scale Natural Images: a database and some
		  statistics},
  booktitle	= {Proceedings of 16'th Danish Conference on Pattern
		  Recognition and Image Analysis (DSAGM) 2008},
  year		= {2008},
  editor	= {Søren I. Olsen},
  series	= {DIKU Technical Report 08-10},
  organization	= {Department of Computer Science, University of Copenhagen,
		  Denmark},
  note		= {Extended abstract},
  project	= {VISIONTRAIN,NISA},
  dikucategory	= {unrefereed}
}

@InProceedings{	  gustavsson.fundana.overgaard.nielsen:07a,
  author	= {David Gustavsson and Ketut Fundana and Niels Chr.
		  Overgaard and Anders Heyden and Mads Nielsen},
  title		= {Variational Segmentation and Contour Matching of Non-Rigid
		  Moving Object},
  booktitle	= {Workshop on Dynamical Vision ({WDV} '07)},
  year		= {2007},
  abstrac	= {In this paper we propose a method for variational
		  segmentation and contour matching of non-rigid objects in
		  image sequences which can deal with the occlusions. The
		  method is based on a region-based active contour model of
		  the Chan-Vese, augmented with a frame-to-frame interaction
		  term which uses the segmentation result from the previous
		  frame as a shape prior. This method has given good results
		  despite the presence of minor occlusions, but can not
		  handle significant occlusions. We have extended this
		  approach by adding a registration step between two
		  consecutive contours. This registration step is based on a
		  novel variational formulation and gives also a mapping of
		  the intensities from the interior of the previous contour
		  to the next. With this information occlusions can be
		  detected from deviations from predicted intensities and the
		  missing intensities in the occluded areas can then be
		  reconstructed. The performance of the method is shown with
		  experiments on synthetic and real image sequences.},
  project	= {VISIONTRAIN}
}

@InProceedings{	  gustavsson.pedersen.nielsen:07a,
  author	= {David Gustavsson and Kim Steenstrup Pedersen and Mads
		  Nielsen},
  title		= {Geometric and Texture Inpainting by Gibbs Sampling},
  booktitle	= {Swedish Symposium on Image Analysis (SSBA '07)},
  year		= {2007},
  month		= mar,
  abstract	= {This paper discuss a method suitable for inpainting both
		  large scale geometric structures and more stochastic
		  texture components. Image inpainting concerns the problem
		  of reconstructing the intensity contents inside regions of
		  missing data. Common techniques for solving this problem
		  are methods based on variational calculus and based on
		  statistical methods. Variationalmethods are good at
		  reconstructing large scale geometric structures but have a
		  tendency to smooth away texture. On the contrary
		  statistical methods can reproduce texture faithfully but
		  fails to reconstruct large scale structures. In this paper
		  we use the well-known FRAME (Filters, Random Fields and
		  Maximum Entropy) for inpainting. We introduce a temperature
		  term in the learned FRAME Gibbs distribution. By sampling
		  using different temperature in the FRAME Gibbs
		  distribution, different contents of the image are
		  reconstructed. We propose a two step method for inpainting
		  using FRAME. First the geometric structure of the image is
		  reconstructed by sampling from a cooled Gibbs distribution,
		  then the stochastic component is reconstructed by sample
		  froma heated Gibbs distribution. Both steps in the
		  reconstruction process are necessary, and contribute in two
		  very different ways to the appearance of the
		  reconstruction.},
  project	= {VISIONTRAIN}
}

@InProceedings{	  gustavsson.pedersen.nielsen:07b,
  author	= {David Gustavsson and Kim Steenstrup Pedersen and Mads
		  Nielsen},
  title		= {Image Inpainting by Cooling and Heating},
  booktitle	= {Scandinavian Conference on Image Analysis ({SCIA} '07)},
  publisher	= {Springer Verlag},
  series	= {Lecture Notes in Computer Science},
  volume	= {4522},
  year		= {2007},
  pages		= {591--600},
  month		= jun,
  editor	= {Bjarne Ersb{\o}ll and Kim Steenstrup Pedersen},
  abstract	= {We discuss a method suitable for inpainting both large
		  scale geometric structures and stochastic texture
		  components. We use the well-known FRAME model for
		  inpainting. We introduce a temperature term in the learnt
		  FRAME Gibbs distribution. By using a fast cooling scheme a
		  MAP-like solution is found that can reconstruct the
		  geometric structure. In a second step a heating scheme is
		  used that reconstruct the stochastic texture. Both steps in
		  the reconstruction process are necessary, and contribute in
		  two very different ways to the appearance of the
		  reconstruction.},
  project	= {VISIONTRAIN}
}