A Robust Multilinear Mixing Model with l2, 1 norm for Unmixing Hyperspectral Images

Minglei Li, Fei Zhu, Alan J. X. Guo

Abstract: The unmixing of hyperspectral data is a hot topic in the field of remote sensing. However, in presence of various types of noise, especially the noisy channels, the performance of unmixing approaches is seriously deteriorated. To enhance the robustness of the unmixing method is a subject worth studying. This paper presents a robust unmixing method based on the recently-proposed multilinear mixing model, where the l2,1 norm is adopted in the loss function to suppress the influence of noise. The sparseness of abundance is also considered to improve the parameter estimation. The resulting optimization problem is solved by the alternating direction multiplier method (ADMM). Experiments on both synthetic and real images demonstrate the performance of the proposed unmixing strategy.

External IDs:dblp:conf/vcip/LiZG20