A Graph Regularized Multilinear Mixing Model for Nonlinear Hyperspectral Unmixing

• Main Authors: Minglei Li, Fei Zhu, Alan J.X. Guo, Jie Chen
• Format: Article
• Language: English
• Published: MDPI AG 2019-09-01
• Series: Remote Sensing
• Subjects: hyperspectral image (HSI); nonlinear unmixing; laplacian graph; alternating direction method of multipliers (ADMM); superpixel
• Online Access: https://www.mdpi.com/2072-4292/11/19/2188

Spectral unmixing of hyperspectral images is an important issue in the fields of remote<br />sensing. Jointly exploring the spectral and spatial information embedded in the data is helpful to<br />enhance the consistency between mixing/unmixing models and real scenarios. This paper proposes<br />a graph regularized nonlinear unmixing method based on the recent multilinear mixing model<br />(MLM). The MLM takes account of all orders of interactions between endmembers, and indicates the<br />pixel-wise nonlinearity with a single probability parameter. By incorporating the Laplacian graph<br />regularizers, the proposed method exploits the underlying manifold structure of the pixels&#8217; spectra,<br />in order to augment the estimations of both abundances and nonlinear probability parameters.<br />Besides the spectrum-based regularizations, the sparsity of abundances is also incorporated for the<br />proposed model. The resulting optimization problem is addressed by using the alternating direction<br />method of multipliers (ADMM), yielding the so-called graph regularized MLM (G-MLM) algorithm.<br />To implement the proposed method on large hypersepectral images in real world, we propose<br />to utilize a superpixel construction approach before unmixing, and then apply G-MLM on each<br />superpixel. The proposed methods achieve superior unmixing performances to state-of-the-art<br />strategies in terms of both abundances and probability parameters, on both synthetic and real datasets.