Abstract
The low spatial resolution of hyperspectral images leads to the coexistence of multiple ground objects in a single pixel (called mixed pixels). A large number of mixed pixels in a hyperspectral image hinders the subsequent analysis and application of the image. In order to solve this problem, a novel sparse unmixing method, which considers highly similar patches in nonlocal regions of a hyperspectral image, is proposed in this article. This method exploits spectral correlation by using collaborative sparsity regularization and spatial information by employing total variation and weighted nonlocal low-rank tensor regularization. To effectively utilize the tensor decomposition, nonlocal similar patches are first grouped together. Then, these nonlocal patches are stacked to form a patch group tensor. Finally, weighted low-rank tensor regularization is enforced to constrain the patch group to obtain an estimated low-rank abundance image. Experiments on simulated and real hyperspectral datasets validated the superiority of the proposed method in better maintaining fine details and obtaining better unmixing results.
Highlights
A hyperspectral image (HSI) is a three-dimensional (3-D) data cube composed of tens or even hundreds of continuous bands, with wavelengths ranging from 400 to 2500 nm
We compare the unmixing performance of the proposed method with the performance of several state-of-the-art methods, such as the CLSUnSAL method [40], SUnSAL with total variation (TV) regularization (SUnSAL-TV) method [45], joint local abundance sparse unmixing (J-LASU) method [60], sparse unmixing with l1-l2 sparsity and TV regularization (l1-l2 SUnSAL-TV) method [74], and the sparse unmixing with nonlocal low-rank prior (NLLRSU) method [75]
We proposed a hyperspectral sparse unmixing method based on weighted nonlocal low-rank tensor decomposition regularization, which takes into account the spatial and spectral information, simultaneously
Summary
A hyperspectral image (HSI) is a three-dimensional (3-D) data cube composed of tens or even hundreds of continuous bands, with wavelengths ranging from 400 to 2500 nm. HSI has a strong nonlocal similarity [63], [64] Based on this prior, a weighted nonlocal low-rank tensor decomposition method for HSI sparse unmixing (WNLTDSU) is proposed in this study. 1) The nonlocal cubic patches are grouped together to form series of low-rank tensors (using weighted nuclear norm) that have potential in exploring the spatial-spectral information deeply and have been proven to perform better than conventional low-rank regularizers in sparse unmixing of HSIs. 2) By utilizing TV, collaborative sparsity and nonlocal tensor low-rank regularizations, the proposed method simultaneously exploits the local spatial smoothness, global row sparsity, and nonlocal low rankness, suppressing noises and maintaining the structural information of the abundance image much better.
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
More From: IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing
Disclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.