TNNG: Total Nuclear Norms of Gradients for Hyperspectral Image Prior

We introduce a novel regularization function for hyperspectral image (HSI), which is based on the nuclear norms of gradient images. Unlike conventional low-rank priors, we achieve gradient-based approximation by minimizing sum associated with rotated planes in HSI. Our method explicitly and simultaneously exploits correlation spectral domain as well spatial domain. low-rankness global region to enhance dimensionality reduction prior. Since our considers domain, it more sensitively detects anomalous variations. achieves high-fidelity recovery using single without explicit use any sparsity-inducing priors such ℓ0, ℓ1 total variation (TV) norms. also apply this robust principal component analysis show its superiority HSI decomposition. To demonstrate, proposed validated variety reconstruction/decomposition problems performance comparisons state-of-the-art methods superior performance.