ADMM-Based Hyperspectral Unmixing Networks for Abundance and Endmember Estimation

Hyperspectral image (HSI) unmixing is an increasingly studied problem in various areas, including remote sensing. It has been tackled using both physical model-based approaches and more recently machine learning-based ones. In this article, we propose a new HSI algorithm combining model- techniques, based on unrolling approaches, delivering improved performance. Our approach unrolls the alternating direction method of multipliers (ADMMs) solver constrained sparse regression underlying linear mixture model. We then neural network structure for abundance estimation that can be trained supervised learning techniques composite loss function. also another blind unsupervised techniques. proposed networks are shown to possess lighter richer containing less learnable parameters skip connections compared with other competing architectures. Extensive experiments show methods achieve much faster convergence better performance even very small training dataset size when methods, such as model-inspired (MNN-AE), (MNN-BU), deep prior (UnDIP), endmember-guided (EGU-Net).