Adaptive spatial regularization for sequential unmixing of hyperspectral images

Abstract

Hyperspectral unmixing is one of the most important tasks in hyperspectral data processing. This work aims to analyze material components coexsiting in low spatial resolution pixels. It is important to integrate both spectral and spatial information to enhance the unmxing performance. Considering that the hyperspectral pixels are usually captured in a sequential manner, an adaptive processing framework is proposed in this paper to efficiently address the unmixing problem with the 11-norm and 12-norm spatial regularization in an online manner. Experiments validated the effectiveness of the proposed methods.