Learning spatial filters for multispectral image segmentation.

Devis Tuia (1) , Rémi Flamary (2) , Gustavo Camps-Valls (1) , Alain Rakotomamonjy (2)

1 IPL - Image Processing Laboratory
2 LITIS - Laboratoire d'Informatique, de Traitement de l'Information et des Systèmes

We present a novel filtering method for multispectral satel- lite image classification. The proposed method learns a set of spatial filters that maximize class separability of binary support vector machine (SVM) through a gradient descent approach. Regularization issues are discussed in detail and a Frobenius-norm regularization is proposed to efficiently exclude uninformative filters coefficients. Experiments car- ried out on multiclass one-against-all classification and tar- get detection show the capabilities of the learned spatial fil- ters.

Mots clés: Image segmentation; Spatial Filtering; Large margin method; multispectral image

Domaines: Apprentissage [cs.LG]

Devis Tuia, Rémi Flamary, Gustavo Camps-Valls, Alain Rakotomamonjy. Learning spatial filters for multispectral image segmentation.. Machine Learning for Signal Processing, Aug 2010, Kittila, Finland. pp.1-6. ⟨hal-00528923⟩