Multiclass ordering for filtering and classification of hyperspectral images

The aim of this paper is to apply genuine hyperspectral mathematical morphology to extract spatial structures according to a set of reference spectra. To achieve this objective, a vector multiclass ordering is introduced in this paper. The proposed ordering is based on a supervised framework which requires a training set. This multiclass supervised ordering may then used for the extension of mathematical morphology to vector images and in particular, we focus here on the application of morphological processing to hyperspectral images, illustrating the performance with real examples. Granulometry for feature description and multi-scale classification are considered as practical application of the proposed approach.