.png){kind=icon}
.png){kind=icon}
Return
A Weighted Fusion Strategy for Robust Endmember Extraction from Hyperspectral Data
.png){kind=icon}
PRE
AI
S
A
DOI：10.1016/j.asr.2026.05.017.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Endmember extraction from hyperspectral images, through the unsupervised identification of materials present in a scene, is a fundamental step in hyperspectral image processing. Existing methods are generally categorized into two main groups: extreme projection-based methods and simplex volume-based methods, each with distinct endmember extraction strategies and varying performance across datasets. To leverage the strengths of both paradigms, this study introduces DFEE, a decision-level fusion-based endmember extraction framework. DFEE integrates endmembers obtained from four well-known, simple, and computationally efficient base algorithms, namely ATGP, VCA, PPI, and N-FINDR, through an unsupervised weighting process guided by image reconstruction error (RE). In essence, DFEE provides a generalized fusion framework that integrates the outputs of these algorithms to achieve an optimal solution, while remaining flexible for implementation with various base methods. Experimental results on both synthetic and real hyperspectral datasets under diverse conditions demonstrate that DFEE consistently outperforms existing methods. Computational analysis further indicates that, due to the efficiency of the base algorithms and the lightweight fusion procedure, DFEE imposes negligible computational overhead.
Keywords:
Endmember extraction
Hyperspectral image processing
Decision-level fusion
Image reconstruction error
Base algorithms
Journal
IF:
2.8
Papers:
1.3K
Citations:
2.0W
Organization
Cited Papers