Hyperspectral Sparse Unmixing With Spectral-Spatial Low-Rank Constraint

Spectral unmixing is a consequential preprocessing task in hyperspectral image interpretation. With the help of large spectral libraries, equivalent to finding optimal subset library entries that can best model image. Sparse regression techniques have been widely used solve this optimization problem, since number materials present scene usually small. However, high mutual coherence signatures negatively affects sparse performance. To cope with challenge, new algorithm called spectral-spatial low-rank (SSLRSU) established. In article, double weighting factors under l 1 framework aim improve row sparsity abundance matrix and each map. Meanwhile, regularization term exploits low-dimensional structure image, which makes for accurate endmember identification from library. The underlying problem be solved by alternating direction method multipliers efficiently. experimental results, conducted using both synthetic real data, uncover proposed SSLRSU strategy get results over those given other advanced strategies.