Summary
Hyperspectral images often have complicated noise that includes random variations and visible patterns like stripes. The authors designed a new denoising method called AXS-Net that breaks down the noisy image into clear parts: the true signal, structured noise, and random noise. Their approach uses a deep learning model inspired by traditional optimization steps, which helps make the results easier to understand and interpret. Tests showed that AXS-Net cleans images well across different types of noise and datasets, and the model’s outputs reveal meaningful components like spectral patterns and noise structures.
hyperspectral imagingimage denoisingstructured noiseGaussian noisedeep unfoldingspectral unmixingproximal operatorslow-rank approximationinterpretable modelsnoise modeling
Abstract
Hyperspectral images (HSIs) are often degraded by mixed noise, including band-dependent Gaussian perturbations and structured artifacts such as stripes, dead-lines, and impulse noise. Most deep denoisers regress the clean image directly, entangling signal and structured noise. We instead model HSI denoising as $\Y=\A\X+\Snoise+\Nnoise$, where $\A\X$ is a low-rank spectral-subspace (unmixing) reconstruction, $\Snoise$ is structured sparse noise and $\Nnoise$ is residual Gaussian noise. The resulting regularized optimization problem is unrolled into AXS-Net, a $K$-stage alternating proximal-point framework. Each stage combines an analytic spectral-basis gradient step, an SSX-Block proximal operator for abundance coefficients, and an SBlock proximal operator for the structured residual with column-consistent and sparse priors. This optimization correspondence exposes interpretable endmembers, abundance maps, and structured-noise estimates. Across ICVL, CAVE, and Harvard datasets and five noise configurations, the proposed AXS-Net achieves strong in-domain accuracy and competitive zero-shot transfer, with consistent gains across all five noise regimes on ICVL and Harvard. The recovered structured-noise closely follows the synthetic reference, and the recovered spectral basis is smooth and band-ordered rather than an arbitrary set of latent channels.