Empirical Inference Technical Report 2010

Cooperative Cuts for Image Segmentation

We propose a novel framework for graph-based cooperative regularization that uses submodular costs on graph edges. We introduce an efficient iterative algorithm to solve the resulting hard discrete optimization problem, and show that it has a guaranteed approximation factor. The edge-submodular formulation is amenable to the same extensions as standard graph cut approaches, and applicable to a range of problems. We apply this method to the image segmentation problem. Specifically, Here, we apply it to introduce a discount for homogeneous boundaries in binary image segmentation on very difficult images, precisely, long thin objects and color and grayscale images with a shading gradient. The experiments show that significant portions of previously truncated objects are now preserved.

Author(s): Jegelka, S. and Bilmes, J.
Number (issue): UWEETR-1020-0003
Year: 2010
Month: August
Day: 0
Bibtex Type: Technical Report (techreport)
Electronic Archiving: grant_archive
Institution: University of Washington, Washington DC, USA
Language: en
Organization: Max-Planck-Gesellschaft
School: Biologische Kybernetik
Links:

BibTex

@techreport{6732,
  title = {Cooperative Cuts for Image Segmentation},
  abstract = {We propose a novel framework for graph-based cooperative regularization that uses submodular costs on graph edges. We introduce an efficient iterative algorithm to solve the resulting hard discrete optimization problem, and show that it has a guaranteed approximation factor. The edge-submodular formulation is amenable to the same extensions as standard graph cut approaches, and applicable to a range of problems. We apply this method to the image segmentation problem. Specifically, Here, we apply it to introduce a discount for homogeneous boundaries in binary image segmentation on very difficult images, precisely, long thin objects and color and grayscale images with a shading gradient. The experiments show that significant portions of previously truncated objects are now preserved.},
  number = {UWEETR-1020-0003},
  organization = {Max-Planck-Gesellschaft},
  institution = {University of Washington, Washington DC, USA},
  school = {Biologische Kybernetik},
  month = aug,
  year = {2010},
  slug = {6732},
  author = {Jegelka, S. and Bilmes, J.},
  month_numeric = {8}
}