Image Segmentation

Segment images with Potts, Metropolis–Hastings, k-means, Otsu, Felzenszwalb, and watershed methods

About this tool

Segment photos locally with several classical algorithms — each in its own tab.

  • Metropolis–Hastings — Monte Carlo Potts-model sampling with adjustable smoothness.
  • Direct Potts — linear-time Q-Ising texture segmentation (Bentrem, arXiv:0906.4745).
  • K-means — Lloyd clustering in RGB color space.
  • Otsu — multi-level histogram thresholding.
  • Felzenszwalb — graph-based region merging.
  • Watershed — immersion on gradient magnitude.

Load an image, pick an algorithm tab, then Preview (fast, downsampled) or Segment (full resolution).