A novel hyperspectral unmixing model based on multilayer NMF with Hoyer's projection

Hyperspectral remote sensing is an important earth observation method with wide application. But the low spatial resolution of hyperspectral images makes it difficult to distinguish the ground objects. The hyperspectral image unmixing is a task to estimate the spectral signatures and corresponding fractional abundances. However, the unmixing speed and efficiency are still limited by traditional structures. In this paper, a novel multilayer nonnegative matrix factorization framework is proposed with Hoyer's projector, called HP-MLNMF. The well-known framework, multilayer nonnegative factorization (MLNMF), is completely restructured and enhanced by introducing the Hoyer's projector to provide the iteration directivity of the structure in the unmixing process. Besides, a novel sparse constraint to spectral signatures suitable for this structure is found as l_1/4-norm based on some experimental discussions. Moreover, the l_p-norm is utilized to find the possible sparest solution for abundance terms. Finally, HP-MLNMF is compared with some representative and state-of-art methods on synthetic and real-world hyperspectral image datasets. Experiments indicate that our method performances well in most cases.