Spectral Variability Aware Blind Hyperspectral Image Unmixing Based on Convex Geometry

Drumetz, Lucas; Chanussot, Jocelyn; Jutten, Christian; Ma, Wing-Kin; Iwasaki, Akira

doi:10.1109/tip.2020.2974062

Cited by 33 publications

(21 citation statements)

References 63 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In low-dimensional models, EV means that the spectral dictionary can vary in space. Taking a step back to HU, we have seen studies that use the matrix factorization method to deal with endmember variability [20][21][22]. In that regard, a possibility one can consider is to adapt such matrix factorization methods to the HSR application.…”

Section: Dictionary-based Regressionmentioning

confidence: 99%

“…As we will see, our global-local low-rank matrix estimation leads to a fairly clean formulation. In comparison, if one applies the EV-present matrix factorization methods in HU, e.g., [22], to HSR, the resulting formulation would be more complicated.…”

Section: Dictionary-based Regressionmentioning

confidence: 99%

“…By definition, a constant L is said to be a Lipschitz constant of ∇g k on X if ∇g k (X)−∇g k (Y ) F ≤ L X − Y F for any X, Y ∈ X . From (22) we have…”

Section: A Proof Of Factmentioning

confidence: 99%

See 2 more Smart Citations

Hyperspectral Super-Resolution via Global–Local Low-Rank Matrix Estimation

et al. 2020

IEEE Trans. Geosci. Remote Sensing

Self Cite

View full text Add to dashboard Cite

Hyperspectral super-resolution (HSR) is a problem that aims to estimate an image of high spectral and spatial resolutions from a pair of co-registered multispectral (MS) and hyperspectral (HS) images, which have coarser spectral and spatial resolutions, respectively. In this paper we pursue a low-rank matrix estimation approach for HSR. We assume that the spectral-spatial matrices associated with the whole image and the local areas of the image have low rank structures. The local low-rank assumption, in particular, has the aim of providing a more flexible model for accounting for local variation effects due to endmember variability. We formulate the HSR problem as a global-local rank-regularized least-squares problem. By leveraging on the recent advances in non-convex large-scale optimization, namely, the smooth Schatten-p approximation and the accelerated majorization-minimization method, we developed an efficient algorithm for the global-local low-rank problem. Numerical experiments on synthetic and semireal data show that the proposed algorithm outperforms a number of benchmark algorithms in terms of recovery performance.

show abstract

Section: Dictionary-based Regressionmentioning

confidence: 99%

Section: Dictionary-based Regressionmentioning

confidence: 99%

See 1 more Smart Citation

Hyperspectral Super-Resolution via Global–Local Low-Rank Matrix Estimation

et al. 2020

IEEE Trans. Geosci. Remote Sensing

Self Cite

View full text Add to dashboard Cite

show abstract

“…In fact, it is wavelength that determines the amount of radiation reflected, scattered, absorbed, or emitted by each material. As a result, spectral signature is highly valued in many real-world applications [3,4], including but not limited to classification [5,6], target detection [7,8], and spectral unmixing [9,10].…”

Section: Introductionmentioning

confidence: 99%

Classification of Remote Sensing Images Through Reweighted Sparse Subspace Representation Using Compressed Data

Zhu¹,

Zhao²,

Wu³

2021

View full text Add to dashboard Cite

In many real-world scenarios, subspace clustering essentially aims to cluster unlabeled high-dimensional data into a union of finite-dimensional linear subspaces. The problem is that the data are always high-dimensional, with the increase of the computation, storge, and communication of various intelligent data-driven systems. This paper attempts to develop a method to cluster spectral images directly using the measurements of compressive coded aperture snapshot spectral imager (CASSI), eliminating the need to reconstruct the entire data cube. Assuming that compressed measurements are drawn from multiple subspaces, a novel algorithm was developed by solving a 1-norm minimization problem, which is known as reweighted sparse subspace clustering (RSSC). The proposed algorithm clusters the compressed measurements into different subspaces, which greatly improves the clustering accuracy over the SSC algorithm by adding a reweighted step. The compressed CASSI measurements obtained using the coherence-based coded aperture can improve the performance of the proposed spectral image clustering method. The accuracy of our spectral image clustering approach was verified through simulations on two real datasets.

show abstract

“…HSIs contain unique rich spectral information. Thus, the pixels can recover the pure spectral feature (endmember) of a series of components via hyperspectral unmixing (HU) and obtain the corresponding proportion (abundance) as well [5], [6].…”

Section: Introductionmentioning

confidence: 99%

Bandwise Model Based on Spectral Prior Information for Sparse Unmixing

Kang

Jiang

et al. 2021

IEEE J. Sel. Top. Appl. Earth Observations Remote Sensing

View full text Add to dashboard Cite

The purpose of sparse unmixing (SU) is to find the optimal spectral subset from the spectral library and uses this subset to model each pixel in the hyperspectral data. The existing SU methods concern Gaussian noise a lot and focus less on the varied intensity of Gaussian noise in different bands and other types of noise, e.g., impulse noise and deadlines. Besides, the high coherence of the spectral library limits the performance of SU. Given the above problems, this paper proposes a new method, called Bandwise Model based on Spectral Prior Information (BMSPI). This proposed BMSPI models the Gaussian noise across different spectral bands and the other types of mixed noise under the maximum a posteriori probability framework, and decreases the effect of high coherence in the spectral library with the spectral prior information. The Alternating Direction Method of Multipliers (ADMM) is adopted to solve the BMSPI. The results of the simulated and real data experiments show that the bandwise model can suppress noises of different types effectively, and the spectral prior information is conducive to guide SU. The advantage of BMSPI is that the mentioned information is used completely. Thus, the accuracy of abundance estimation is improved.

show abstract

Spectral Variability Aware Blind Hyperspectral Image Unmixing Based on Convex Geometry

Cited by 33 publications

References 63 publications

Hyperspectral Super-Resolution via Global–Local Low-Rank Matrix Estimation

Hyperspectral Super-Resolution via Global–Local Low-Rank Matrix Estimation

Classification of Remote Sensing Images Through Reweighted Sparse Subspace Representation Using Compressed Data

Bandwise Model Based on Spectral Prior Information for Sparse Unmixing

Contact Info

Product

Resources

About