Sparse hyperspectral unmixing via arctan approximation of L0 norm

Yaser Esmaeili Salehani, Saeed Gazor, Il-Min Kim, Shahram Yousefi

Abstract: In this paper, we introduce a method of hyperspectral unmixing in the linear mixing model with the given library of the constituent materials. The proposed algorithm employs an arctan function to approximate the l0 norm in the minimization problem. This approximation makes the objective function smooth, facilitates the convergence and results in reduced reconstruction errors. We evaluate the proposed method and compare it with other methods via simulation. This reveals that the proposed method outperforms the state-of-the-art methods and results in higher reconstruction signal-to-noise-ratio.