Image Segmentation with Adaptive Sparse Grids

Benjamin Peherstorfer, Julius Adorf, Dirk Pflüger, Hans-Joachim Bungartz

Abstract

We present a novel adaptive sparse grid method for unsupervised image segmentation. The method is based on spectral clustering. The use of adaptive sparse grids achieves that the dimensions of the involved eigensystem do not depend on the number of pixels. In contrast to classical spectral clustering, our sparse-grid variant is therefore able to segment larger images. We evaluate the method on real-world images from the Berkeley Segmentation Dataset. The results indicate that images with 150,000 pixels can be segmented by solving an eigenvalue system of dimensions 500 × 500 instead of 150, 000 × 150, 000.
BibTeX
@inproceedings{peherstorfer2013image,
  author = {Benjamin Peherstorfer and Julius Adorf and Dirk Pflüger and Hans-Joachim Bungartz},
  title = {{Image Segmentation with Adaptive Sparse Grids}},
  booktitle = {Australasian Conference on Artificial Intelligence},
  year = {2013},
  doi = {10.1007/978-3-319-03680-9\_17},
}

Loading related content...

Figure