From Proper Scoring Rules and Edge Biases to Mumford-Shah Models: Robust Unsupervised Hyperspectral Image Segmentation

Key points

• Hyperspectral images provide a rich representation of the underlying spectrum for each pixel, allowing for a pixel-wise classification/segmentation into different classes.1
• As the availability of labeled hyperspectral images is limited and the acquisition of labeled training data is time-consuming, unsupervised segmentation methods are important in hyperspectral image segmentation.1
• Our framework turns a strictly proper scoring rule and a family of parametric probability distributions into an indicator function for the Mumford-Shah functional, and we naturally generalize some previous Mumford-Shah based models for image segmentation this way.1
• Our focus will be on heavy-tailed elliptical distributions, the negative log-likelihood score and the score based on the beta-divergence to deal with the heavy outliers in hyperspectral imaging.1

Sources (1)

• [1]From Proper Scoring Rules and Edge Biases to Mumford-Shah Models: Robust Unsupervised Hyperspectral Image Segmentation
  arXiv (AI, ML, NLP, CV, robotics, multi-agent) · Oct 20, 10:04 AM

  We present a feature-agnostic, unsupervised hyperspectral segmentation framework based on the piecewise-constant Mumford-Shah functional.

  Hyperspectral images provide a rich representation of the underlying spectrum for each pixel, allowing for a pixel-wise classification/segmentation into different classes.

Extractive summary: sentences quoted from the sources.