Hyperspectral Unmixing Hierarchies

arXiv cs.CV / 4/21/2026

📰 NewsIdeas & Deep AnalysisModels & Research

共有:

Key Points

The paper tackles key challenges in hyperspectral unmixing, including spectral variability, ambiguity in choosing the number of endmembers, and endmember degradation as more are included.
It proposes hierarchical unmixing by enforcing a hierarchical abundance sum constraint within Deep Nonnegative Matrix Factorization, and introduces a compact model called Binary Linear Unmixing Tactile Hierarchies (BLUTHs).
The authors further refine BLUTHs per-scene using sparsity modulation, aiming to better match each scene’s topology and improve robustness to spectral variability.
Results show BLUTHs outperform state-of-the-art methods on laboratory hyperspectral scenes (notably for abundance estimation), remain competitive on remote sensing data, and are demonstrated for ocean color unmixing on HYPSO and PACE satellite imagery.
Overall, the hierarchical structure is positioned as a unifying approach that helps recover endmembers with different spectral contrast levels more clearly than conventional unmixing pipelines.

Abstract

Unmixing reveals the spatial distribution and spectral details of different constituents, called endmembers, in a hyperspectral image. Because unmixing has limited ground truth requirements, can accommodate mixed pixels, and is closely tied to light propagation, it is a uniquely powerful tool for analyzing hyperspectral images. However, spectral variability inhibits unmixing performance, the proper way to determine the number of endmembers is ambiguous, and the clarity of the endmembers degrades as more are included. Hierarchical structure is a possible solution to all three problems. Here, hierarchical unmixing is defined by imposing a hierarchical abundance sum constraint on Deep Nonnegative Matrix Factorization. Binary Linear Unmixing Tactile Hierarchies (BLUTHs) solve the hierarchical unmixing problem with a simple network architecture. Sparsity modulation unmixing growth tailors the topology of a BLUTH to each scene. The structure imposed by BLUTHs allows endmembers with varying levels of spectral contrast to be revealed, mitigating the challenge of spectral variability. The performance of BLUTHs exceeds state-of-the-art unmixing algorithms on laboratory scenes, particularly with regard to abundance estimation, while their performance remains competitive on remote sensing scenes. In addition, ocean color unmixing by BLUTHs is demonstrated on hyperspectral scenes from the HYPSO and PACE satellites.