Nonlinear Unmixing for Hyperspectral Images via Kernel-Transformed Bilinear Mixing Models

Gu, Jiafeng; Yang, Bin; Wang, Bin

doi:10.1109/tgrs.2021.3135571

Cited by 8 publications

(4 citation statements)

References 51 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…, An [26]. Hyperspectral unmixing can be mainly classified into two, Linear mixture-based models [12], [27] and nonlinear mixture based models [26], [28], [29]. Endmember extraction is the process of identifying the pure pixels (also known as pure signatures) of a specific class.…”

Section: Hyperspectral Unmixingmentioning

confidence: 99%

Blind nonlinear unmixing using nonnegative matrix factorization based bi-objective autoencoder

Bhakthan,

Loganathan,

Bansal

2023

IJEECS

View full text Add to dashboard Cite

Hyperspectral image processing is one of the trending techniques used in many fields such as remote sensing, medical, agriculture, food processing, and military. The unique discrimination of hyperspectral images can be used for object identification, classification, and prediction. One of the main challenges of these tasks is the mixed pixel problem. Hyperspectral unmixing is the process of identifying the endmembers and their abundance in pixels. In linear unmixing, the mixture of the endmembers is assumed to be linear homogenous patches. Even though these models are simple and faster in performance, most of the real-world images are not linear. A modified nonlinear mixture-based sparsity regularized bi-objective autoencoder model based on nonnegative matrix factorization (NMF-BOA) is proposed in this article. The performance analysis shows that our model gives competitive results compared to the state-of-the-art models.

show abstract

Section: Hyperspectral Unmixingmentioning

confidence: 99%

Blind nonlinear unmixing using nonnegative matrix factorization based bi-objective autoencoder

Bhakthan,

Loganathan,

Bansal

2023

IJEECS

View full text Add to dashboard Cite

show abstract

“…We are interested in characterizing the specific contribution of a fluorophore of interest. Blind unmixing algorithms are generic and can be applied to various domains [1], [10], [13], [14] However, in our case, because they have limitations because they cannot accurately isolate contributions that are specific to a particular fluorophore of interest.…”

Section: Introductionmentioning

confidence: 99%

An Explicit Estimated Baseline Model for Robust Estimation of Fluorophores Using Multiple-Wavelength Excitation Fluorescence Spectroscopy

Gautheron

Sdika

Hébert

et al. 2024

IEEE Trans. Biomed. Eng.

View full text Add to dashboard Cite

Spectroscopy is a popular technique for identifying and quantifying fluorophores in fluorescent materials. However,quantifying the fluorophore of interest can be challenging when the material also contains other fluorophores (baseline), particularly if the emission spectrum of the baseline is not well-defined and overlaps with that of the fluorophore of interest. In this work, we propose a method that is free from any prior assumptions about the baseline by utilizing fluorescence signals at multiple excitation wavelengths. Despite the nonlinearity of the model, a closed-form expression of the least squares estimator is also derived.To evaluate our method, we consider the practical case of estimating the contributions of two forms of protoporphyrin IX (PpIX) in a fluorescence signal. This fluorophore of interest is commonly utilized in neuro-oncology operating rooms to distinguish the boundary between healthy and tumor tissue in a type of brain tumor known as glioma. Using a digital phantom calibrated with clinical and experimental data, we demonstrate that our method is more robust than current state-of-the-art methods for classifying pathological status, particularly when applied to images of simulated clinical gliomas. To account for the high variability in the baseline, we are examining various scenarios and their corresponding outcomes. In particular, it maintains the ability to distinguish between healthy and tumor tissue with an accuracy of up to 87%, while the ability of existing methods drops near 0%.

show abstract

“…The p-order polynomial model [15] and the multilinear mixing (MLM) model [16] account for high-order interactions among endmembers. Recently, considerable endeavors have been made to enhance NMM-based unmixing, including the analysis of band-wise nonlinearity [17], kernel-transformed BMMs [18], and robustness to complex noise [19], etc. In addition, the use of deep learning (DL) techniques, such as autoencoders, has also contributed to nonlinear unmixing.…”

Section: Introductionmentioning

confidence: 99%

Unsupervised Nonlinear Hyperspectral Unmixing with Reduced Spectral Variability via Superpixel-Based Fisher Transformation

Yin,

Yang

2023

Remote Sensing

Self Cite

View full text Add to dashboard Cite

In hyperspectral unmixing, dealing with nonlinear mixing effects and spectral variability (SV) is a significant challenge. Traditional linear unmixing can be seriously deteriorated by the coupled residuals of nonlinearity and SV in remote sensing scenarios. For the simplification of calculation, current unmixing studies usually separate the consideration of nonlinearity and SV. As a result, errors individually caused by the nonlinearity or SV still persist, potentially leading to overfitting and the decreased accuracy of estimated endmembers and abundances. In this paper, a novel unsupervised nonlinear unmixing method accounting for SV is proposed. First, an improved Fisher transformation scheme is constructed by combining an abundance-driven dynamic classification strategy with superpixel segmentation. It can enlarge the differences between different types of pixels and reduce the differences between pixels corresponding to the same class, thereby reducing the influence of SV. Besides, spectral similarity can be well maintained in local homogeneous regions. Second, the polynomial postnonlinear model is employed to represent observed pixels and explain nonlinear components. Regularized by a Fisher transformation operator and abundances’ spatial smoothness, data reconstruction errors in the original spectral space and the transformed space are weighed to derive the unmixing problem. Finally, this problem is solved by a dimensional division-based particle swarm optimization algorithm to produce accurate unmixing results. Extensive experiments on synthetic and real hyperspectral remote sensing data demonstrate the superiority of the proposed method in comparison with state-of-the-art approaches.

show abstract

Nonlinear Unmixing for Hyperspectral Images via Kernel-Transformed Bilinear Mixing Models

Cited by 8 publications

References 51 publications

Blind nonlinear unmixing using nonnegative matrix factorization based bi-objective autoencoder

Blind nonlinear unmixing using nonnegative matrix factorization based bi-objective autoencoder

An Explicit Estimated Baseline Model for Robust Estimation of Fluorophores Using Multiple-Wavelength Excitation Fluorescence Spectroscopy

Unsupervised Nonlinear Hyperspectral Unmixing with Reduced Spectral Variability via Superpixel-Based Fisher Transformation

Contact Info

Product

Resources

About