Multi-feature Hyperspectral Image Classification with Local and Non-local Spatial Information via Markov Random Field in Semantic Space

Xiangrong Zhang, Zeyu Gao, Licheng Jiao, Huiyu Zhou

Research output: Contribution to journalArticle

18 Citations (Scopus)

Abstract

Hyperspectral images (HSIs) provide invaluable information in both spectral and spatial domains for image classification tasks. In this paper, we use semantic representation as a middle-level feature to describe image pixels' characteristics. Deriving effective semantic representation is critical for achieving good classification performance. Since different image descriptors depict characteristics from different perspectives, combining multiple features in the same semantic space makes semantic representation more meaningful. First, a probabilistic support vector machine is used to generate semantic representation-based multifeatures. In order to derive better semantic representation, we introduce a new adaptive spatial regularizer that well exploits the local spatial information, while a nonlocal regularizer is also used to search for global patch-pair similarities in the whole image. We combine multiple features with local and nonlocal spatial constraints using an extended Markov random field model in the semantic space. Experimental results on three hyperspectral data sets show that the proposed method provides better performance than several state-of-the-art techniques in terms of region uniformity, overall accuracy, average accuracy, and Kappa statistics.
Original languageEnglish
JournalIEEE Trans. on Geoscience and Remote Sensing
Early online date09 Nov 2017
DOIs
Publication statusEarly online date - 09 Nov 2017

Fingerprint

Image classification
image classification
Semantics
pixel
Support vector machines
Pixels
Statistics

Cite this

@article{4d6bda94cacb4d7fb5346a683b296d3d,
title = "Multi-feature Hyperspectral Image Classification with Local and Non-local Spatial Information via Markov Random Field in Semantic Space",
abstract = "Hyperspectral images (HSIs) provide invaluable information in both spectral and spatial domains for image classification tasks. In this paper, we use semantic representation as a middle-level feature to describe image pixels' characteristics. Deriving effective semantic representation is critical for achieving good classification performance. Since different image descriptors depict characteristics from different perspectives, combining multiple features in the same semantic space makes semantic representation more meaningful. First, a probabilistic support vector machine is used to generate semantic representation-based multifeatures. In order to derive better semantic representation, we introduce a new adaptive spatial regularizer that well exploits the local spatial information, while a nonlocal regularizer is also used to search for global patch-pair similarities in the whole image. We combine multiple features with local and nonlocal spatial constraints using an extended Markov random field model in the semantic space. Experimental results on three hyperspectral data sets show that the proposed method provides better performance than several state-of-the-art techniques in terms of region uniformity, overall accuracy, average accuracy, and Kappa statistics.",
author = "Xiangrong Zhang and Zeyu Gao and Licheng Jiao and Huiyu Zhou",
year = "2017",
month = "11",
day = "9",
doi = "10.1109/TGRS.2017.2762593",
language = "English",
journal = "IEEE Trans. on Geoscience and Remote Sensing",
issn = "0196-2892",
publisher = "Institute of Electrical and Electronics Engineers Inc.",

}

Multi-feature Hyperspectral Image Classification with Local and Non-local Spatial Information via Markov Random Field in Semantic Space. / Zhang, Xiangrong; Gao, Zeyu; Jiao, Licheng; Zhou, Huiyu.

In: IEEE Trans. on Geoscience and Remote Sensing, 09.11.2017.

Research output: Contribution to journalArticle

TY - JOUR

T1 - Multi-feature Hyperspectral Image Classification with Local and Non-local Spatial Information via Markov Random Field in Semantic Space

AU - Zhang, Xiangrong

AU - Gao, Zeyu

AU - Jiao, Licheng

AU - Zhou, Huiyu

PY - 2017/11/9

Y1 - 2017/11/9

N2 - Hyperspectral images (HSIs) provide invaluable information in both spectral and spatial domains for image classification tasks. In this paper, we use semantic representation as a middle-level feature to describe image pixels' characteristics. Deriving effective semantic representation is critical for achieving good classification performance. Since different image descriptors depict characteristics from different perspectives, combining multiple features in the same semantic space makes semantic representation more meaningful. First, a probabilistic support vector machine is used to generate semantic representation-based multifeatures. In order to derive better semantic representation, we introduce a new adaptive spatial regularizer that well exploits the local spatial information, while a nonlocal regularizer is also used to search for global patch-pair similarities in the whole image. We combine multiple features with local and nonlocal spatial constraints using an extended Markov random field model in the semantic space. Experimental results on three hyperspectral data sets show that the proposed method provides better performance than several state-of-the-art techniques in terms of region uniformity, overall accuracy, average accuracy, and Kappa statistics.

AB - Hyperspectral images (HSIs) provide invaluable information in both spectral and spatial domains for image classification tasks. In this paper, we use semantic representation as a middle-level feature to describe image pixels' characteristics. Deriving effective semantic representation is critical for achieving good classification performance. Since different image descriptors depict characteristics from different perspectives, combining multiple features in the same semantic space makes semantic representation more meaningful. First, a probabilistic support vector machine is used to generate semantic representation-based multifeatures. In order to derive better semantic representation, we introduce a new adaptive spatial regularizer that well exploits the local spatial information, while a nonlocal regularizer is also used to search for global patch-pair similarities in the whole image. We combine multiple features with local and nonlocal spatial constraints using an extended Markov random field model in the semantic space. Experimental results on three hyperspectral data sets show that the proposed method provides better performance than several state-of-the-art techniques in terms of region uniformity, overall accuracy, average accuracy, and Kappa statistics.

U2 - 10.1109/TGRS.2017.2762593

DO - 10.1109/TGRS.2017.2762593

M3 - Article

JO - IEEE Trans. on Geoscience and Remote Sensing

JF - IEEE Trans. on Geoscience and Remote Sensing

SN - 0196-2892

ER -