Using a Panchromatic Image to Improve Hyperspectral Unmixing

摘要: Hyperspectral unmixing is a widely studied field of research aiming at estimating the pure material signatures and their abundance fractions from hyperspectral images. Most spectral methods are based on prior knowledge assumptions that induce limitations, such as existence least one pixel for each material. This work presents new approach to overcome some these limitations by introducing co-registered panchromatic image in process. Our method, called Heterogeneity-Based Endmember Extraction coupled with Local Constrained Non-negative Matrix Factorization (HBEE-LCNMF), has several steps: first set endmembers estimated heterogeneity criterion applied followed clustering. Then, order complete this endmember set, local using constrained non-negative matrix factorization strategy, proposed. The performance our regards criteria, compared those state-of-the-art obtained synthetic satellite data describing urban periurban scenes, considering French HYPXIM/HYPEX2 mission characteristics. images built real reflectances do not contain endmember. simulated airborne acquisition spatial features mission. method demonstrates benefit reduce well-known data. On data, reduces angle between spectra 46% Vertex Component Analysis (VCA) N-finder (N-FINDR) methods. HBEE-LCNMF other yield equivalent performance, but, proposed shows more robustness over sets tested Moreover, does require know number endmembers.