#### DMCA

## Gabor-Filtering-Based Nearest Regularized Subspace for Hyperspectral Image Classification

Citations: | 2 - 1 self |

### Citations

931 | Robust face recognition via sparse representation
- Wright, Yang, et al.
- 2009
(Show Context)
Citation Context ...ace. Some commonly implemented kernel functions for SVM include the polynomial kernel and the radial-basis-function (RBF) kernel [16], [17]. Recently, sparse-representation-based classification (SRC) =-=[18]-=- was developed for pattern classification, originally for face recognition. SRC classifier is essentially based on the concept that a pixel can be represented as a sparse linear combination of labeled... |

776 |
Uncertainty relation for resolution in space, spatial frequency, and orientation optimized by two-dimensional visual cortical filters
- Daugman
- 1985
(Show Context)
Citation Context ...OR-FILTERING-BASED NRS A. Gabor Filter AGabor filter,1 which is a sinusoidal function modulated by a Gaussian envelope, has been popularly used in applications of computer vision and image processing =-=[32]-=-, [33]. In a twodimensional ( , ) coordinate system, the Gabor filter, including a real component and imaginary one, can be represented as where where represents the wavelength of the sinusoidal facto... |

107 | Sparse representation or collaborative representation: Which helps face recognition
- Zhang, Yang, et al.
(Show Context)
Citation Context ...each class independently. The -minimization is imposed due to the argument that it was the collaborative representation rather than sparse representation playing the essential role for classification =-=[23]-=-. The essence of NRS classifier is an penalty, however, in the style of a distance-weighted Tikhonov regularization [24]. On the one hand, rather than enforcing a strong assumption about the nature of... |

80 | Spectral and Spatial Classification of Hyperspectral Data Using SVMs and Morphological Profiles.
- Fauvel
- 2008
(Show Context)
Citation Context ...ent from Bayes classifiers, some classifiers can work well in HSI even without a presupposed data distribution and a preprocess of dimensionality reduction. For instance, support vector machine (SVM) =-=[13]-=-–[15] is a state-of-the-art classifier which has shown superior performance for hyperspectral classification tasks, especially under the SSS condition. SVM classifier seeks to separate classes by lear... |

66 | Joint Bayesian endmember extraction and linear unmixing for hyperspectral imagery
- Dobigeon, Moussaoui, et al.
(Show Context)
Citation Context ...ades, hyperspectral classification has been an important application of HSI, including urban growth monitoring, biological and chemical detection, environmental monitoring, mineral exploration, etc., =-=[2]-=-–[4]. A Bayes classifier, such as maximum-likelihood-estimation (MLE) [5] or Gaussian mixture model (GMM) [6], is one of the most commonly used classifiers in HSI. Due to the high dimensionality, line... |

53 |
Regularization studies of linear discriminant analysis in small sample size scenarios with application to face recognition,
- Lu, Plataniotis, et al.
- 2005
(Show Context)
Citation Context ...or HSI when the number of available samples is even smaller than the dimensionality. A few solutions have been proposed for the SSS problem associated with LDA. One solution is regularized LDA (RLDA) =-=[9]-=-, [10], for which amatrix with small energy on the diagonals is added to the ill-conditioned scatter matrix to stabilize the estimation of the LDA projection. The other popular approach is to use subs... |

42 | SVM- and MRF-based method for accurate classification of hyperspectral images
- Tarabalka, Fauvel, et al.
- 2010
(Show Context)
Citation Context ...rom Bayes classifiers, some classifiers can work well in HSI even without a presupposed data distribution and a preprocess of dimensionality reduction. For instance, support vector machine (SVM) [13]–=-=[15]-=- is a state-of-the-art classifier which has shown superior performance for hyperspectral classification tasks, especially under the SSS condition. SVM classifier seeks to separate classes by learning ... |

29 | Hyperspectral image classification using dictionary-based sparse representation
- Chen, Nasrabadi, et al.
- 2011
(Show Context)
Citation Context ...n, originally for face recognition. SRC classifier is essentially based on the concept that a pixel can be represented as a sparse linear combination of labeled samples via the -norm minimization. In =-=[19]-=-, the authors introduced sparse-based representation for hyperspectral image classification; in [20], the authors further presented the sparse-based -minimization for hyperspectral target detection wh... |

27 |
Classification of hyperspectral images with regularized linear discriminant analysis
- Bandos, Bruzzone, et al.
- 2009
(Show Context)
Citation Context ...I when the number of available samples is even smaller than the dimensionality. A few solutions have been proposed for the SSS problem associated with LDA. One solution is regularized LDA (RLDA) [9], =-=[10]-=-, for which amatrix with small energy on the diagonals is added to the ill-conditioned scatter matrix to stabilize the estimation of the LDA projection. The other popular approach is to use subspace-b... |

25 |
An efficient discriminant-based solution for small sample size problem,”
- Das, Nenadic
- 2009
(Show Context)
Citation Context ...minant analysis (LFDA) [7], are usually applied to reduce the dimensionality for hyperspectral data and generally provide good classification results. However, under small-sample-size (SSS) condition =-=[8]-=-, LDA and its variants potentially fail owing to ill-conditioned statistical estimates particularly when the dimensionality of the input feature space is very high, as is the case for HSI when the num... |

24 |
Similarity-based unsupervised band selection for hyperspectral image analysis
- Du, Yang
- 2008
(Show Context)
Citation Context ...er is first considered to exploit useful spatial information in a PCA-projected subspace.Moreover, we also discuss the implementation of Gabor filter in a subset of original bands with band selection =-=[34]-=-. In [34], linear prediction error has been proposed for unsupervised band selection based on band similarity measurement. In this work, Gabor features are employed as the preprocessing of NRS classif... |

18 | Locality-preserving dimensionality reduction and classification for hyperspectral image analysis
- Li, Prasad, et al.
- 2012
(Show Context)
Citation Context ... [6], is one of the most commonly used classifiers in HSI. Due to the high dimensionality, linear discriminant analysis (LDA) [5] and its variants, such as local Fisher’s discriminant analysis (LFDA) =-=[7]-=-, are usually applied to reduce the dimensionality for hyperspectral data and generally provide good classification results. However, under small-sample-size (SSS) condition [8], LDA and its variants ... |

17 | Semisupervised one-class support vector machines for classification of remote sensing data
- Mũnoz-Marí, Bovolo, et al.
- 2010
(Show Context)
Citation Context ... the original space to linearly separable ones in the kernel-induced space. Some commonly implemented kernel functions for SVM include the polynomial kernel and the radial-basis-function (RBF) kernel =-=[16]-=-, [17]. Recently, sparse-representation-based classification (SRC) [18] was developed for pattern classification, originally for face recognition. SRC classifier is essentially based on the concept th... |

14 |
Structured gaussian components for hyperspectral image classification,” Geoscience and Remote Sensing
- Berge, Solberg
- 2006
(Show Context)
Citation Context ...g, biological and chemical detection, environmental monitoring, mineral exploration, etc., [2]–[4]. A Bayes classifier, such as maximum-likelihood-estimation (MLE) [5] or Gaussian mixture model (GMM) =-=[6]-=-, is one of the most commonly used classifiers in HSI. Due to the high dimensionality, linear discriminant analysis (LDA) [5] and its variants, such as local Fisher’s discriminant analysis (LFDA) [7],... |

11 | 2011b, Sparse representation for target detection in hyperspectral imagery
- CHEN, NASRABADI, et al.
(Show Context)
Citation Context ...can be represented as a sparse linear combination of labeled samples via the -norm minimization. In [19], the authors introduced sparse-based representation for hyperspectral image classification; in =-=[20]-=-, the authors further presented the sparse-based -minimization for hyperspectral target detection which requires a priori information, such as available labeled samples. In [21], the sparse logistic r... |

11 |
Hyperspectral image classification based on structured sparse logistic regression and 3d wavelet texture features
- Qian, Ye
- 2013
(Show Context)
Citation Context ...age classification; in [20], the authors further presented the sparse-based -minimization for hyperspectral target detection which requires a priori information, such as available labeled samples. In =-=[21]-=-, the sparse logistic regression was discussed in a three-dimensional wavelet domain for hyperspectral classification. Note that such representation-based classification does not Manuscript received M... |

11 | Nearest regularized subspace for hyperspectral classification
- Li, Tramel, et al.
- 2014
(Show Context)
Citation Context ...ing pixel) is represented by labeled samples (or training samples), and the class label to be assigned will be the same one as that of labeled samples providing the lowest representation residual. In =-=[22]-=-, we developed nearest regularized subspace (NRS) classifier, for which an approximation of each sample to be classified is represented via a linear combination of all available labeled samples of eac... |

11 |
On combining multiple features for hyperspectral remote sensing image classification
- Zhang, Zhang, et al.
- 2012
(Show Context)
Citation Context ... Spatial information has been verified to be helpful for hyperspectral image classification [25], [26]. Recently,Gabor features have been successfully used for hyperspectral image classification [27]–=-=[30]-=- due to the ability to represent useful spatial information. In [27] and [28], three-dimensional Gabor feature extractionmethods were discussed; in [29] and [30], twodimensional Gabor features were ex... |

9 | An adaptive mean-shift analysis approach for object extraction and classification from urban hyperspectral imagery,” Geoscience and Remote Sensing - Huang, Zhang - 2008 |

9 |
An SVM ensemble approach combining spectral, structural, and semantic features for the classification of high-resolution remotely sensed imagery
- Huang, Zhang
(Show Context)
Citation Context ...ring locations. Yet, for HSI, it is highly probable that two adjacent pixels belong to the same class. Spatial information has been verified to be helpful for hyperspectral image classification [25], =-=[26]-=-. Recently,Gabor features have been successfully used for hyperspectral image classification [27]–[30] due to the ability to represent useful spatial information. In [27] and [28], three-dimensional G... |

8 |
Three-Dimensional Gabor Wavelets for Pixel-Based Hyperspectral Imagery Classification
- Shen, Jia
- 2011
(Show Context)
Citation Context ...classification [25], [26]. Recently,Gabor features have been successfully used for hyperspectral image classification [27]–[30] due to the ability to represent useful spatial information. In [27] and =-=[28]-=-, three-dimensional Gabor feature extractionmethods were discussed; in [29] and [30], twodimensional Gabor features were extracted in a PCA-projected subspace, and fed into some classifiers, such as S... |

8 |
A comparative study of spatial approaches for urban mapping using hyperspectral ROSIS images over Pavia City
- Huang, Zhang
- 2009
(Show Context)
Citation Context ... Spectrometer (ROSIS) sensor. The image scene, with a spatial coverage of pixels covering the city of Pavia, Italy, was collected under the HySens project managed by DLR (the German Aerospace Agency) =-=[35]-=-. The dataset has 103 spectral bands prior to water-band removal, with a spectral coverage from 0.43 to and a spatial resolution of 1.3 m. Approximately, 42 776 labeled pixels with nine classes are fr... |

7 |
Signal Theory Methods
- Landgrebe
- 2003
(Show Context)
Citation Context ...RODUCTION A CQUIRED by remote-sensing systems (e.g., spaceborneor airborne sensors), hyperspectral imagery (HSI) typically is represented as a three-dimensional cubewith hundreds of spectral channels =-=[1]-=-. Each hyperspectral pixel records reflectance over these contiguous and narrow spectral bands. This allows for distinguishing or classifying materials with subtle changes in their reflectance signatu... |

5 | Joint within-class collaborative representation for hyperspectral image classification
- Li, Du
- 2014
(Show Context)
Citation Context ...echnique in the first category. Specifically, we extract Gabor features, which can be implemented very easily. Note that some recent investigation of techniques in the second category can be found in =-=[31]-=-; for SRC, the condition of local continuity intends to choose the same labeled sample for the representation of neighboring pixels; however, due to the collaborative (not competitive) nature among th... |

4 |
Target detection under misspecified models in hyperspectral images
- Bajorski
- 2012
(Show Context)
Citation Context ..., hyperspectral classification has been an important application of HSI, including urban growth monitoring, biological and chemical detection, environmental monitoring, mineral exploration, etc., [2]–=-=[4]-=-. A Bayes classifier, such as maximum-likelihood-estimation (MLE) [5] or Gaussian mixture model (GMM) [6], is one of the most commonly used classifiers in HSI. Due to the high dimensionality, linear d... |

4 |
Segmentation of hyperspectral images via subtractive clustering and cluster validation using one-class support vector machines. Geoscience and Remote Sensing
- Bilgin, Erturk, et al.
- 2011
(Show Context)
Citation Context ...riginal space to linearly separable ones in the kernel-induced space. Some commonly implemented kernel functions for SVM include the polynomial kernel and the radial-basis-function (RBF) kernel [16], =-=[17]-=-. Recently, sparse-representation-based classification (SRC) [18] was developed for pattern classification, originally for face recognition. SRC classifier is essentially based on the concept that a p... |

3 |
Object-oriented subspace analysis for airborne hyperspectral remote sensing imagery
- Zhang, Huang
- 2010
(Show Context)
Citation Context ...eighboring locations. Yet, for HSI, it is highly probable that two adjacent pixels belong to the same class. Spatial information has been verified to be helpful for hyperspectral image classification =-=[25]-=-, [26]. Recently,Gabor features have been successfully used for hyperspectral image classification [27]–[30] due to the ability to represent useful spatial information. In [27] and [28], three-dimensi... |

3 |
Hyperspectral region classification using a three-dimensional Gabor filterbank
- Bau, Sarkar, et al.
(Show Context)
Citation Context ...lass. Spatial information has been verified to be helpful for hyperspectral image classification [25], [26]. Recently,Gabor features have been successfully used for hyperspectral image classification =-=[27]-=-–[30] due to the ability to represent useful spatial information. In [27] and [28], three-dimensional Gabor feature extractionmethods were discussed; in [29] and [30], twodimensional Gabor features we... |

2 |
Dimensionality Reduction and Linear Discriminant Analysis for Hyperspectral Image Classification
- Du, Younan
(Show Context)
Citation Context ...s added to the ill-conditioned scatter matrix to stabilize the estimation of the LDA projection. The other popular approach is to use subspace-based dimensionality reduction as a preprocessing to LDA =-=[11]-=-, [12], where, typically, principal component analysis (PCA) [5] isfirst employed to reduce the dimensionality of the input data discarding the null space of rank-deficient scatter matrices so that th... |

2 | Noise-adjusted subspace discriminant analysis for hyperspectral image classification,” IEEEGeosci
- Li, Prasad, et al.
- 2013
(Show Context)
Citation Context ...d to the ill-conditioned scatter matrix to stabilize the estimation of the LDA projection. The other popular approach is to use subspace-based dimensionality reduction as a preprocessing to LDA [11], =-=[12]-=-, where, typically, principal component analysis (PCA) [5] isfirst employed to reduce the dimensionality of the input data discarding the null space of rank-deficient scatter matrices so that the LDA ... |

2 |
Spectral and Spatial Classification of Hyperspectral Data Using SVMs and Gabor Textures
- Huo, Tang
- 2011
(Show Context)
Citation Context ...sed for hyperspectral image classification [27]–[30] due to the ability to represent useful spatial information. In [27] and [28], three-dimensional Gabor feature extractionmethods were discussed; in =-=[29]-=- and [30], twodimensional Gabor features were extracted in a PCA-projected subspace, and fed into some classifiers, such as SVM. In this paper, we investigate the benefits of using Gabor features for ... |

1 | Classification of hyperspectral images by using extended morphological attribute profiles and independent component analysis - DallaMura, Benediktsson, et al. - 2011 |

1 |
DesigningGaborfilters for optimal texture separability,”Pattern Recognit
- Jernigan
- 2000
(Show Context)
Citation Context ...TERING-BASED NRS A. Gabor Filter AGabor filter,1 which is a sinusoidal function modulated by a Gaussian envelope, has been popularly used in applications of computer vision and image processing [32], =-=[33]-=-. In a twodimensional ( , ) coordinate system, the Gabor filter, including a real component and imaginary one, can be represented as where where represents the wavelength of the sinusoidal factor, rep... |