DESTRIPING HYPERSPECTRAL IMAGERY BY ADAPTIVE ANISOTROPIC TOTAL VARIATION AND TRUNCATED NUCLEAR NORM

Abstract

Destriping is an important hyperspectral image processing procedure. Many variation-based methods show adequate destriping performance. However, global variation constraints cause loss of texture information in unstriped image regions. To alleviate this effect, we propose an adaptive anisotropy total variation method to adaptively smoothen the striped regions. Furthermore, considering the highly linear relationship among stripes, we introduce, for the first time, the truncated nuclear norm to constrain the rank of the stripes to 1. When combining the adaptive anisotropy total variation and the truncated nuclear norm, a hyperspectral image destriping model is established, which can easily be solved by the alternating direction method of multipliers (ADMM). Experiments demonstrate the effectiveness and superiority of the proposed destriping method.