Blind Hyperspectral Unmixing With Integrated Nonsmooth and CEM Spatial Constraints

Read full text Summarise with VIRAT
Authors: Shaoquan Zhang, Kexing Li, Xinyi Zhou, Fan Li, Pengfei Lai, Lianhui Liang, Chengzhi Deng, Shengqian Wang
Cited by: 0
References: 45

Summary

No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.

Summarise with VIRAT Free to try with your Science Catalogs account.

Abstract

Hyperspectral blind unmixing serves as a key technique in remote sensing image processing. Its core value lies in the ability to extract endmember spectral signatures and invert abundance fractions from mixed pixels without relying on prior endmember information. However, in practical applications, prevalent noise contamination in hyperspectral data often leads to degraded signal-to-noise ratio (SNR), which gives rise to spectral distortion and diminished image quality, consequently imposing a fundamental limitation on the accuracy of spectral unmixing. To overcome these challenges, this study introduces an innovative unmixing algorithm under the$\ell _{1}$-NMF framework that integrates endmember smoothness constraints with spatial abundance regularization based on Constrained Energy Minimization (CEM). Specifically, a non-smooth matrix is incorporated to enforce strong sparsity on endmembers, simultaneously smoothing image noise and improving endmember estimation accuracy. Furthermore, a novel dual-weight sparse regularization term is constructed by combining a CEM detector with a spatial weight, which concurrently enhances both the sparse representation and spatial continuity of abundance maps. On synthetic and real hyperspectral datasets, the proposed EISNMF algorithm yields greater accuracy in estimating both endmembers and abundances compared to existing blind unmixing techniques, proving to be particularly robust against high levels of noise.

Fields of Study & Subjects

Remote-Sensing Image Classification Spectroscopy and Chemometric Analyses Advanced Image Fusion Techniques

Publication details

Journal / source
IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing
Volume
19
Pages
12134–12151
Publication date
1 January 2026
Language
EN