SNMF-Net: Learning a Deep Alternating Neural Network for Hyperspectral Unmixing

Hyperspectral unmixing is recognized as an important tool to learn the constituent materials and corresponding distribution in a scene. The physical spectral mixture model always tackle this problem because of its highly ill-posed nature. In article, we introduce linear (LMM)-based end-to-end deep neural network named SNMF-Net for hyperspectral unmixing. shares alternating architecture benefits from both model-based methods learning-based methods. On one hand, high interpretability it built by unrolling $L_{p}$ sparsity constrained nonnegative matrix factorization ( -NMF) belonging LMM families. other all parameters submodules can be seamlessly linked with optimization algorithm -NMF problem. This enables us reasonably integrate prior knowledge on unmixing, algorithm, sparse representation theory into robust learning, so improve Experimental results synthetic real-world data show advantages proposed over many state-of-the-art