Abstract:Sparse unmixing aims at finding an optimal subset of spectral signatures in a large spectral library to effectively model each pixel in the hyperspectral image and compute their fractional abundances. In most previous work concerned with the sparse unmixing, L2 norm is used to measure the error tolerance and the L1 norm is added as the sparsity regularization. However, in some applications, using L1 norm to measure the error tolerance has significant robustness advantages over the L2 norm. Besides, in some cases, using a smooth function to approximate the L0 norm can obtain more accurate results than the L1 norm in the field of sparse regression. Thus, in this paper, we consider the two alternative choices for sparse unmixing. A reweighted iteration algorithm is also proposed so that the unconvex regularizer (smoothed L0 norm) can be efficiently solved through transforming it into a series of weighted L1 regularizer problems. Experimental results on both synthetic and real hyperspectral data demonstrate the efficacy of the new models. The work was supported by the National Natural Science Foundation of China under the Grants 61273245 and 91120301, the 973 Program under the Grant 2010CB327904, the open funding project of State Key Laboratory of Virtual Reality Technology and Systems, Beihang University (Grant No. BUAA-VR-12KF-07), and Program for New Century Excellent Talents in University of Ministry of Education of China under the Grant NCET-11-0775 and the Beijing Natural Science Foundation (Non-negative Component Analysis for Hyperspectral Imagery Unmixing) under the Grant 4112036.The work was also supported by Beijing Key Laboratory of Digital Media, Beihang University, Beijing 100191, P.R. China. ∗Corresponding author. Tel.: +86-10-823-39-520; Fax: +86-10-823-38-798. Email address: shizhenwei@buaa.edu.cn (Zhenwei Shi) Preprint submitted to Elsevier January 27, 2013

Spectral Variation Alleviation by Low-Rank Matrix Approximation for Hyperspectral Image Analysis.

Hyperspectral Image Classification by Exploring Low-Rank Property in Spectral or/and Spatial Domain.

Hyperspectral Anomaly Detection Via Locally Enhanced Low-Rank Prior

How to Fully Explore the Low-Rank Property for Data Recovery of Hyperspectral Images

Simultaneous Spatial and Spectral Low-Rank Representation of Hyperspectral Images for Classification

Enhanced Tensor Low-Rank Representation Learning for Hyperspectral Anomaly Detection

A Fast Low Rank Approximation and Sparsity Representation Approach to Hyperspectral Anomaly Detection

Hyperspectral Anomaly Detection Based on Low-Rank Representation Using Local Outlier Factor

Improved Low-Rank Matrix Approximation for Hyperspectral Image Spatial-Spectral Feature Extraction

Adaptive $textit{L}_{bf 1/2}$ Sparsity-Constrained NMF With Half-Thresholding Algorithm for Hyperspectral Unmixing

Hyperspectral Anomaly Detection Via $S_{1/2}$ and Total Variation Low Rank Matrix Decomposition

Hyperspectral Anomaly Detection Based on Adaptive Low-Rank Transformed Tensor

Hyperspectral Image Denoising Via Spectral and Spatial Low-Rank Approximation

An Improved Low Rank and Sparse Matrix Decomposition-Based Anomaly Target Detection Algorithm for Hyperspectral Imagery

Projection Subspace Based Low-Rank Representation for Sparse Hyperspectral Unmixing

Sparse Hyperspectral Unmixing Using An Approximate L 0

Simultaneously Sparse and Low-Rank Abundance Matrix Estimation for Hyperspectral Image Unmixing

Denoising of Hyperspectral Images Using Nonconvex Low Rank Matrix Approximation.

Low-rank and Spectral-Spatial Variation Regularized Hyperspectral Image Denoising Algorithm

Sparse Hyperspectral Unmixing Using an Approximate L0 Norm

Hyperspectral Anomaly Detection Via Merging Total Variation into Low-Rank Representation