Adaptive Total Variation Regularization for Weighted Low-Rank Tensor Sparse Hyperspectral Unmixing

Hyperspectral images have a low spatial resolution due to the limitations of satellite imaging equipment, resulting in multiple substances contained in a pixel (mixed pixels). The phenomenon of mixed pixels affects the subsequent analyses and researches on hyperspectral images. To address this problem, we propose a novel hyperspectral unmixing method named adaptive total variation regularization for weighted low-rank tensor sparse hyperspectral unmixing (ATVWLRTSU). This method considers the spatial structural characteristics of different regions in the hyperspectral image by using the adaptive Total Variation (ATV) term, and exploits the abundance low-rank tensor by utilizing weighted nuclear norm in hyperspectral unmixing. Simulation and real experiments show that the proposed method has better performance in terms of both anti-noise and details. © (2024), (International Association of Engineers). All rights reserved.