Spatial and spectral segmentation of satellite remote sensing imagery using processing graphs by mathematical morphology

The paper presents an autoadaptive method of satellite images segmentation based on a description of the image by graphs. The method presented proposes a new approach using the morphological analysis of the dual graph of the segmentation. A quantification builds an image oversegmented in a connected components set forming a new digital grid. Morphological operations are applied to the graph corresponding to form significant segments by region growing. In the process of growth, the attributes of the regions are kept to enslave the process, but also to contribute to the interpretation of the relations of neighborhood and to understand the physics of detection used.