International Journal of Information Technology & Computer Science ( IJITCS )
Locally preserving projection (LPP) does not take advantage of the spatial correlation of pixels in the image, and the pixels are considered as independent pieces of information. In this paper, a kernel based manifold learning feature extraction method which considers spatial relationship of neighboring pixels, called supervised composite kernel locality preserving projection (SCKLPP), is proposed for hyperspectral image feature extraction. The spatial information and spectral information from original hyperspectral image are combined using composite kernels weight matrix. The nearest neighbor graph is created with the prior class-label information of samples. Experimental results on AVIRIS data set show that the SCKLPP can not only efficiently reduce the dimensionality but also achieve higher accuracies. In addition, the proposed method opens a new field for future developments in which spatial information can be easily integrated into the feature extraction stage..
: feature extraction; dimensionality reduction; locally preserving projection; composite kernel; hyperspectral image classification
 L. O. Jimenez, J. L. Rivera-median, E. Rodirguez-diaz, et al. “Integration of spatial and spectral information homgenous by means of unsupervised extraction and classification for objects applied to multispectral and hyperspectral data”, IEEE Transaction on Geoscience and Remote Sensing, 43, pp. 844-851, 2005.
 C. M. Bachmann, T. L. Ainsworth and R. A. Fusina. “Exploiting manifold geometry in hyperspectral imagery”, IEEE Transaction on Geoscience and Remote Sensing, 43, pp.441-454, 2005.
 T. Han, and G. Goodenough. “Nonlinear feature extraction of hyperspectral data based on Locally Linear Embedding (LLE)”, IEEE Geoscience and Remote Sensing Symposium, pp. 1237-1240, 2005.
 Y. C. Chen, M. M. Crawford and J. Ghosh. “Applying nonlinear manifold learning to hyperspectral data for land cover classification”, IEEE Geoscience and Remote Sensing Symposium, pp. 4311-4314, 2005.
 A. Mohan, G. Sapiro and E. Bosch. “Spatially coherent nonlinear dimensionality reduction and segmentation of hyperspectral images”, IEEE Geoscience and Remote Sensing Letters, 4, pp. 206-210, 2007.
 G. Y. Chen, and S. E. Qian. “Dimensionality reduction of hyperspectral imagery using improved locally linear embedding”, Journal of Applied Remote Sensing, 1, 2007.
 S. E. Qian, and G. Y. Chen. “A new nonlinear dimensionality reduction method with application to hyperspectral image analysis”, IEEE International Geoscience and Remote Sensing Symposium, pp. 270 273, 2007.
 X. F. He and P. Niyogi. “Locality preserving projections”, Advances in Neural Information Processing System 16, MIT Press, Cambridge, MA, pp. 153-160, 2003.
 J. Cheng, Q. Shan, H. Lu et al. “Supervised kernel locality preserving projections for face recognition”, Neurocomputing, 67, pp. 443-44, 2005.
 X. Huang, L. Zhang. “A comparative study of spatial approaches for urban mapping using hyperspectral ROSIS images over Pavia city, northern of Italy”, International Journal of Remote Sensing, 30, pp. 3205-3221, 2009.
 L. P. Zhang, X. Huang. “Object-oriented subspace analysis for airborne hyperspectral remote sensing imagery”, Neurocomputing, 73, pp. 927936, 2010. G. Camps-Valls, L. Gomez-chova. “Composite kernels for hyperspectral image classification”, IEEE Transaction on Geoscience and Remote Sensing Letters, 3, pp. 93-97, 2006.
 V. F. Santiago, V. Manian. “Improving hyperspectral image classification using spatial preprocessing”, IEEE Transaction on Geoscience and Remote Sensing Letters, 6, 297-301, 2009.
 M. Belkin and P. Niyogi. “Laplacian eigenmaps and spectral techniques for embedding and clustering”, In Advances in Neural Information Processing System, 14, Cambridge, MA, USA, pp. 585-591, 2002.
 V. F. Santiago, J. Angulo. “Morphological scale-space for hyperspectral image and dimensionality exploration using tensor modeling”, Hyperspectral Image and Signal Processing: Evolution in Remote Sensing, 2009, WHISPERS '09, 1-4