Image analysis of hyperspectral data using mathematical morphology

The purpose of this tutorial is to get familiar with some techniques for the analysis of remote sensing hyperspectral images exploiting the spatial information of the image. When dealing with hyperspectral images with high spatial resolution, the spatial relations of the pixels in the scene are fundamental for the analysis. Classical techniques for hyperspectral images addressing different tasks (e.g., classification, object extraction and change detection) that consider only the spectral characteristics of the pixels are limited since this complementary information source is not properly exploited. Different approaches exist for including the spatial information in the analysis (e.g., post-processing based on spatial regularization, use of spatial features and segmentation). In this tutorial we will focus on the extraction and use of spatial features for hyperspectral image analysis. In particular considering features based on operators defined in the mathematical morphology framework.