Hyperspectral unmixing by reweighted low rank and total variation

In recent years, sparse regression has drawn much attention in hyperspectral unmixing. The well known sparse unmixing via variable splitting augmented Lagrangian (SUnSAL) and sparse unmixing via variable splitting augmented Lagrangian and total variation (SUnSAl-TV) aim to find the sparsest abundance of every data vector individually. However, these methods ignore the global structure of all the vectors. In this paper, we propose a novel hyperspectral unmixing method by exploiting low rank property of the abundance matrix. Our proposed method find the lowest-rank representation of a collection of the abundance vectors by using reweighted low rank constraint. This way, our proposed unmixing method better captures the global structure of the abundance matrix and improve the accuracy of abundance estimation. Our approach also takes the spatial context into account by a TV constraint. Experimental results on both the synthetic and real hyperspectral data demonstrate the effectiveness of our proposed algorithm