Collaborative Sparse Hyperspectral Unmixing Using $l_{0}$ Norm

Zhenwei Shi,Tianyang Shi,Min Zhou,Xia Xu

DOI: https://doi.org/10.1109/tgrs.2018.2818703

IF: 8.2

2018-01-01

IEEE Transactions on Geoscience and Remote Sensing

Abstract:Sparse unmixing has been applied on hyperspectral imagery popularly in recent years. It assumes that every observed signature is a linear combination of just a few spectra (end-members) from a known spectral library. However, solving the sparse unmixing problem directly (using l(0) norm to control the sparsity of solution at a low level) is NP-hard. Most related works focus on convex relaxation methods, but the sparsity and accuracy of results cannot be well guaranteed. Under these circumstances, this paper proposes a novel algorithm termed collaborative sparse hyperspectral unmixing using l(0) norm (CSUnL0), which aims at solving l(0) problem directly. First, it introduces a row-hard-threshold function. The row-hard-threshold function makes it possible to combine l(0) norm, instead of its approximate norms, with alternating direction method of multipliers. Compared with the convex relaxation methods, the l(0) norm constraint guarantees sparser and more accurate results. Moreover, the antinoise ability of CSUnL0 also gets improved. Second, CSUnL0 uses l(2) norm of each end-members' abundance across the whole map as a collaborative constraint, which can take advantage of the hyperspectral data's subspace property. The experimental results indicate that l(0) norm contributes to acquiring a more sparser solution and helps CSUnL0 to enhance calculation accuracy.