Sparse Nonnegative Matrix Factorization for Hyperspectral Unmixing Based on Endmember Independence and Spatial Weighted Abundance

Hyperspectral image unmixing is an important task for remote sensing processing. It aims at decomposing the mixed pixel of to identify a set constituent materials called endmembers and obtain their proportions named abundances. Recently, number algorithms based on sparse nonnegative matrix factorization (NMF) have been widely used in hyperspectral with good performance. However, these NMF only consider correlation characteristics abundance usually just take Euclidean structure data into account, which can make extracted become inaccurate. Therefore, aim addressing this problem, we present algorithm endmember independence spatial weighted paper. Firstly, it assumed that should be independent from each other. Thus, by utilizing autocorrelation endmembers, constraint constructed model. In addition, two weights neighborhood pixels coefficient are proposed estimated smoother so as further explore underlying data. The not considers relevant abundances simultaneously, but also makes full use spatial-spectral information image, achieving more desired experiment results several sets verify effectiveness algorithm.