Exploiting the low-rank property of hyperspectral imagery: A technical overview

Abstract

Hyperspectral images (HSIs) often suffer from various annoying degradations, which poses huge challenges for the practical applications. Fortunately, clean HSI is intrinsically low-rank, which opens up a broad category of HSI processing and analysis methods with high robustness against the complicated mixture of various noises and outliers. Based on the low rank property of HSI, this paper provides a comprehensive review on restoration, multiangle registration and unmixing methods for HSIs developed very recently, and insights for further investigations.