Hyperspectral Image Clustering with Spatially-Regularized Ultrametrics

TL;DR

This paper introduces a novel unsupervised hyperspectral image clustering method that leverages spatially regularized ultrametric spectral clustering, combining data density and geometry for accurate, label-free material classification with theoretical guarantees.

Contribution

It presents a new clustering approach that efficiently integrates spatial regularization and ultrametric distances, providing robust performance guarantees and automatic cluster number estimation.

Findings

Achieves high labeling accuracy on synthetic and real data

Efficient with quasilinear scaling in data size

Successfully estimates the number of clusters automatically

Abstract

We propose a method for the unsupervised clustering of hyperspectral images based on spatially regularized spectral clustering with ultrametric path distances. The proposed method efficiently combines data density and geometry to distinguish between material classes in the data, without the need for training labels. The proposed method is efficient, with quasilinear scaling in the number of data points, and enjoys robust theoretical performance guarantees. Extensive experiments on synthetic and real HSI data demonstrate its strong performance compared to benchmark and state-of-the-art methods. In particular, the proposed method achieves not only excellent labeling accuracy, but also efficiently estimates the number of clusters.