Looking into a Pixel by Nonlinear Unmixing -- A Generative Approach

2026-04-01

AI Summary

This work addresses the nonlinear spectral mixing problem in hyperspectral images, which arises due to large pixel footprints, by proposing a model-free, invertible unmixing framework that does not rely on explicit spectral mixture models. The method introduces generative modeling into hyperspectral nonlinear unmixing for the first time, constructing a CycleGAN-based LCGU-Net architecture that leverages cycle-consistency constraints and a linear–nonlinear mixing association mechanism to achieve end-to-end unsupervised unmixing. Experimental results demonstrate that the proposed approach consistently outperforms state-of-the-art model-based methods across multiple datasets, significantly enhancing the generalizability and robustness of nonlinear unmixing.

Abstract

Due to the large footprint of pixels in remote sensing imagery, hyperspectral unmixing (HU) has become an important and necessary procedure in hyperspectral image analysis. Traditional HU methods rely on a prior spectral mixing model, especially for nonlinear mixtures, which has largely limited the performance and generalization capacity of the unmixing approach. In this paper, we address the challenging problem of hyperspectral nonlinear unmixing (HNU) without explicit knowledge of the mixing model. Inspired by the principle of generative models, where images of the same distribution can be generated as that of the training images without knowing the exact probability distribution function of the image, we develop an invertible mixing-unmixing process via a bi-directional GAN framework, constrained by both the cycle consistency and the linkage between linear and nonlinear mixtures. The combination of cycle consistency and linear linkage provides powerful constraints without requiring an explicit mixing model. We refer to the proposed approach as the linearly-constrained CycleGAN unmixing net, or LCGU net. Experimental results indicate that the proposed LCGU net exhibits stable and competitive performance across different datasets compared with other state-of-the-art model-based HNU methods.

Problem

• hyperspectral unmixing
• nonlinear unmixing
• spectral mixing model
• remote sensing
• model-free unmixing

Innovation

• hyperspectral nonlinear unmixing
• generative model
• CycleGAN
• model-free unmixing
• invertible mixing-unmixing