MCGSMs Applied to Multiscale Image Representations


msmcgsm.tar.gz (243.6 MB)


The archive contains code for training mixtures of conditional Gaussian scale mixtures on multiscale representations of natural images. The archive also contains already preprocessed training data extracted from the van Hateren dataset. For more information, read this paper or the README file contained in the archive.


  author = "L. Theis and R. Hosseini and M. Bethge",
  title = "Mixtures of Conditional Gaussian Scale Mixtures Applied to Multiscale Image Representations",
  year = 2012,
  journal = "PLoS ONE",
  publisher = "Public Library of Science",
  volume = 7,
  number = 7,
  month = "Jul",
  keywords = "natural image statistics, gaussian scale mixtures, random fields",
  doi = "10.1371/journal.pone.0039857"
University of Tuebingen BCCN CIN MPI