Geometric Spatio-Spectral Total Variation for Hyperspectral Image Denoising and Destriping

Abstract

This article proposes a novel regularization method, named Geometric Spatio-Spectral Total Variation (GeoSSTV), for hyperspectral (HS) image denoising and destriping. HS images are inevitably affected by various types of noise due to the measurement equipment and environment. Total Variation (TV)-based regularization methods that model the spatio-spectral piecewise smoothness inherent in HS images are promising approaches for HS image denoising and destriping. However, existing TV-based methods are based on classical anisotropic and isotropic TVs, which cause staircase artifacts and lack rotation invariance, respectively, making it difficult to accurately recover round structures and oblique edges. To address this issue, GeoSSTV introduces a geometrically consistent formulation of TV that measures variations across all directions in a Euclidean manner. Through this formulation, GeoSSTV removes noise while preserving round structures and oblique edges. Furthermore, we formulate the HS image denoising problem as a constrained convex optimization problem involving GeoSSTV and develop an efficient algorithm based on a preconditioned primal-dual splitting method. Experimental results on HS images contaminated with mixed noise demonstrate the superiority of the proposed method over existing approaches.