Hyperspectral Texture Classification Using a Markov Model and a Projection Pursuit Technique. Classification de Textures Hyperspectrales Fondée sur un Modèle Markovien et une Technique de Poursuite de Projection

Hyperspectral Texture Classification Using a Markov Model and a Projection Pursuit Technique

Classification de Textures Hyperspectrales Fondée sur un Modèle Markovien et une Technique de Poursuite de Projection

Guillaume Rellier Xavier Descombes  Frédéric Falzon  Josiane Zerubia 

Page: 
25-42
|
Received: 
18 January 2003
|
Accepted: 
N/A
|
Published: 
31 March 2003
| Citation

OPEN ACCESS

Abstract: 

In this paper we tackle the problem of hyperspectral texture analysis. In order to take advantage of the images fine spectral sampling, we perform a joint spatial and spectral texture modeling using a vectorial approach within a Markovian framework. This model is used jointly with a projection pursuit algorithm, for the determination of an optimal subspace on which the data are projected. This is done to limit the effect of the high dimensionality of the data (known as Hughes phenomenon, or curse of dimensionality). This model is tested for urban areas supervised classification. 

Résumé

Nous considérons le problème de l'analyse de textures d'images hyperspectrales. De manière à tirer parti de la finesse de la discrétisation spectrale inhérente à ce type d'image, nous réalisons une modélisation de texture qui intègre simultanément les données spatiales et spectrales en employant une approche vectorielle dans un modèle markovien. Ce modèle est utilisé conjointement avec un algorithme de poursuite de projection, permettant de déterminer un sous-espace optimal dans lequel on projette les données. Ceci permet de diminuer les effets néfastes d'une trop grande dimensionnalité des données, connus sous le nom de phénomène de Hughes. Ce modèle est testé dans le cadre de la classification supervisée de zones urbaines. 

Keywords: 

Hyperspectral image, classification, texture, dimension reduction, projection pursuit, Markov random fields.

Mots clés

Image hyperspectrale, classification, texture, réduction de la dimension, poursuite de projection, champs de Markov.  

1. introduction
2. Réduction de la Dimension
3. Modélisation
4. Application de la Modélisation Proposée à la Réduction de la Dimension de l'Espace pour la Classification
5. Résultats
6. Conclusion
  References

[Bel61] R. Bellman, Adaptive control processes : a guided tour. Princeton Univ. Press., 1961. 

[Bes74] J. Besag, Spatial interaction and the statistical analysis of lattice systems. J. Roy. Stat. Soc. B, 36:192-236, 1974. 

[BK99] J. Bennett et A. Khotanzad, Maximum likelihood estimation methods for multispectral random field image models. IEEE Transactions on Pattern Analysis and Machine Intelligence (PAMI), 21(6):537-543, 1999. 

[CCB85] R. Chellappa, S. Chatterjee, et R. Bagdazian, Texture synthesis and compression using Gauss-Markov random fields. IEEE Transactions on Systems, Man, and Cybernetics, 8(2):298-303, Mars-Avril 1985. 

[DE87] H. Derin et H. Elliott, Modeling and segmentation of noisy and textured images using Gibbs random fields. IEEE Transactions on Pattern Analysis and Machine Intelligence (PAMI), 9:721-741, 1987. 

[Des93] X. Descombes, Champs markoviens en analyse d'images. Thèse de Doctorat, ENST Paris, 1993. 

[Des97] X. Descombes, A dense class of Markov Random Fields and associated parameter estimation. Journal of Visual Communication and Image Representation, 8(3):299-316, Septembre 1997.

[DMZ97] X. Descombes, R. Morris, et J. Zerubia, Quelques améliorations à la segmentation d'image bayésienne. Traitement du Signal, 14(4):373-382, 1997. 

[DSP99] X. Descombes, M. Sigelle, et F. Préteux, Estimating Gaussian Markov random field parameters in a nonstationary framework : application to remote sensing imaging. IEEE Transactions on Image Processing (IP), 8(4):490-503, 1999. 

[Dun89] G. H. Dunteman, Principal Components Analysis. Sage Inc., 1989. 

[Fuk90] K. Fukunaga, Introduction to statistical pattern recognition. Academic Press, 2e edition, 1990. 

[GBSC88] A.A. Green, M. Berman, P. Switzer, et M.D. Craig, A transformation for ordering multispectral data in terms of image quality with implications for noise removal. IEEE Transactions on Geoscience and Remote Sensing (GRS), 26(1):65-74, 1988.

[GG84] S. Geman et D. Geman, Stochastic relaxation, Gibbs distribution and the Bayesian restauration of images. IEEE Transactions on Pattern Analysis and Machine Intelligence (PAMI), 6(6):721-741, Novembre 1984. 

[Gim99] G.L. Gimel'farb, Image texture and Gibbs random fields. Kluwer, 1999. 

[Har79] R. M. Haralick, Statistical and structural approaches to texture. Proceedings of the IEEE, 67(5):786-804, 1979. 

[Haz00] G. G. Hazel, Multivariate Gaussian MRF for multispectral scene segmentation and anomaly detection. IEEE Transactions on Geoscience and Remote Sensing (GRS), 38(3):1199-1211, 2000.

[HL93] P. Hall et K. Li, On almost linearity of low dimensionality projections from high dimensional data. The Annals of Statistics, 21(2):867-889, 1993. 

[Hub85] P. J. Huber, Projection pursuit. The Annals of Statistics, 13(2):435-475, 1985. 

[Hug68] G.F. Hughes, On the mean accuracy of statistical pattern recognizers. IEEE Transactions on Infirmation Theory, 14(1):55-63, 1968. 

[Jai89] A. K. Jain, Fundamentals of image processing, chapter 2 : Two-dimensional systems and mathematical preliminaries. Prentice Hall, 1989. 

[JF91] A. K. Jain et F. Farrokhnia, Unsupervised texture segmentation using Gabor filters. Pattern Recognition, 24(12):1167-1186, 1991. 

[JL98] L.O. Jimenez et D.A. Landgrebe, Supervised classification in highdimensional space : geometrical, statistical and asymptotical properties of multivariate data. IEEE Transactions on Systems, Man and CyberneticsPart C : Applications and Reviews, 28(1):39-54, Février 1998. 

[JL99] L. O. Jimenez et D. A. Landgrebe, Hyperspectral data analysis and supervised feature reduction via projection pursuit. IEEE Transactions on Geoscience and Remote Sensing (GRS), 37(6):2653-2667, 1999. 

[JR99] X. Jia et J. A. Richards, Segmented principal components transformation for efficient hyperspectral remote-sensing image display and classification. IEEE Transactions on Geoscience and Remote Sensing (GRS), 37(1):538-542, 1999. 

[KH95] C. Kervrann et F. Heitz, A Markov random field model-based approach to unsupervised texture segmentation using local and global spatial statistics. IEEE Transactions on Image Processing (IP), 4(6):856-862, Juin 1995. 

[LC00] C. Lee et E. Choi, Bayes error evaluation of the Gaussian ML classifier. IEEE Transactions on Geoscience and Remote Sensing (GRS), 38(3): 1471-1475, 2000. 

[LDZ00] A. Lorette, X. Descombes, et J. Zerubia, Urban areas extraction based on texture analysis through a Markovian modelling. International Journal of Computer Vision (IJCV), 36(3):219-234, 2000. 

[PH95] D. K. Panjwani et G. Healey, Markov random field models for unsupervised segmentation of textured color images. IEEE Transactions on Pattern Analysis and Machine Intelligence (PAMI), 17(10):939-954, 1995. 

[SM00] S. M. Schweizer et J. M. F. Moura, Hyperspectral imagery : Clutter adaptation in anomaly detection. IEEE Transactions on Infirmation Theory, 46(5):1855-1871, 2000. 

[Win95] G. Winkler, Image analysis, random fields and dynamic Monte Carlo methods : a mathematical introduction. Springer, 1995. 

[Wu82] F.Y Wu, The Potts model. Reviews of modern physics, 54(1):235-267, 1982.