High-Dimensional Noise to Low-Dimensional Manifolds: A Manifold-Space Diffusion Framework for Robust Degraded Hyperspectral Classification
Abstract:
In recent years, Hyperspectral Image (HSI) classification has attracted increasing attention in remote sensing due to its critical role in fine-grained land-cover mapping and semantic understanding.However, HSI data are inherently high-dimensional but low-rank, with discriminative information concentrated on a low-dimensional latent manifold. In real-world remote sensing scenarios, the superposition of multiple degradation factors disrupts this intrinsic manifold structure, driving samples away from their original low-dimensional distribution and introducing substantial redundant and non-discriminative variations.To better handle this challenge, this paper proposes a manifold-space diffusion framework for robust hyperspectral classification under complex degradation conditions. Specifically, the proposed method first embeds high-dimensional, degradation-corrupted hyperspectral observations into a compact low-dimensional manifold by introducing a discriminatively guided spectral–spatial reconstruction task, which preserves class-discriminative semantics while suppressing redundant and non-informative variations. Building upon this manifold representation, a diffusion-based generative model is further employed to explicitly model and regularize the spectral–spatial distribution in the manifold space, enabling progressive refinement and stabilization of the latent features against residual composite degradations.The key advantage of the proposed framework lies in performing diffusion-based distribution modeling directly on the low-dimensional manifold, effectively decoupling degradation-induced disturbances from intrinsic discriminative structures and enhancing representation stability under complex degradations.Experimental results on multiple hyperspectral benchmarks demonstrate consistent performance improvements over state-of-the-art methods under diverse composite degradation settings.