Abstract:Hyperspectral unmixing is an important processing step for many hyperspectral applications, mainly including: 1) estimation of pure spectral signatures (endmembers) and 2) estimation of the abundance of each endmember in each pixel of the image. In recent years, nonnegative matrix factorization (NMF) has been highly attractive for this purpose due to the nonnegativity constraint that is often imposed in the abundance estimation step. However, most of the existing NMF-based methods only consider the information in a single layer while neglecting the hierarchical features with hidden information. To alleviate such limitation, in this paper, we propose a new sparsity-constrained deep NMF with total variation (SDNMF-TV) technique for hyperspectral unmixing. First, by adopting the concept of deep learning, the NMF algorithm is extended to deep NMF model. The proposed model consists of pretraining stage and fine-tuning stage, where the former pretrains all factors layer by layer and the latter is used to reduce the total reconstruction error. Second, in order to exploit adequately the spectral and spatial information included in the original hyperspectral image, we enforce two constraints on the abundance matrix. Specifically, the $L_{1/2}$ constraint is adopted, since the distribution of each endmember is sparse in the 2-D space. The TV regularizer is further introduced to promote piecewise smoothness in abundance maps. For the optimization of the proposed model, multiplicative update rules are derived using the gradient descent method. The effectiveness and superiority of the SDNMF-TV algorithm are demonstrated by comparing with other unmixing methods on both synthetic and real data sets.

Combined Nonlocal Spatial Information and Spatial Group Sparsity in NMF for Hyperspectral Unmixing

Hypergraph-Regularized Sparse NMF for Hyperspectral Unmixing.

Kernel Based Sparse NMF Algorithm for Hyperspectral Unmixing

Spectral–Spatial Joint Sparse NMF for Hyperspectral Unmixing

Nonlocal Low-Rank Nonnegative Tensor Factorization for Hyperspectral Unmixing

Superpixel-Based Nonnegative Tensor Factorization for Hyperspectral Unmixing

Spectral–Spatial Robust Nonnegative Matrix Factorization for Hyperspectral Unmixing

Spectral-Spatial Hyperspectral Unmixing Using Nonnegative Matrix Factorization

Subspace Structure Regularized Nonnegative Matrix Factorization for Hyperspectral Unmixing

Nonnegative Matrix Factorization with Data-Guided Constraints for Hyperspectral Unmixing.

L1/2 Sparsity Constrained Nonnegative Matrix Factorization for Hyperspectral Unmixing

Hyperspectral Unmixing Using Sparsity-Constrained Deep Nonnegative Matrix Factorization With Total Variation

Nonnegative matrix factorization with region sparsity learning for hyperspectral unmixing

Nonlinear Hyperspectral Unmixing Based on Constrained Multiple Kernel NMF

A Kernel Spatial Complexity-Based Nonlinear Unmixing Method of Hyperspectral Imagery.

Spatial Group Sparsity Regularized Nonnegative Matrix Factorization for Hyperspectral Unmixing

Nonnegative matrix factorization with piecewise smoothness constraint for hyperspectral unmixing

Nonlinear Hyperspectral Unmixing Based on Sparse Non-Negative Matrix Factorization

Hyperspectral Unmixing Based on Nonnegative Matrix Factorization: A Comprehensive Review