Sparse hyperspectral unmixing combined L1/2 norm and reweighted total variation regularization

Sparse regression aims at estimating the fractional abundances of pure endmembers based on the assumption that each mixed pixel in the hyperspectral image can be expressed in the form of linear combinations of a number of known and pure endmembers. And total variation spatial regularization for sparse unmixing has been proposed with incorporating spatial information. In this paper, considering the desirable performance of reweighted minimization and owing to the L-1/2 norm is an alternative regularizer which is much easier to solved than L-0 regularizer and has better sparsity and robustness than L-1 regularizer, a sparse regression combined L-1/2 norm and reweighted total variation regularization has been utilized. Then the unconvex optimization problem is simply solved by the variable splitting and augmented Lagrangian algorithm. Our experimental results with simulated data sets and real hyperspectral data sets demonstrate that the proposed method is an effective and accurate spectral unmixing algorithm for hyperspectral regression.