Mathematical Morphology applied to Very High Resolution Spatial images interpretation

Nowadays, the emergence of very high resolution satellite (VHRS) images imposes to reconsider the methods used to describe surfaces and objects in remote sensing images. The precision provided by new sensors allow the capability to discriminate small entities and non-observable objects until now particularly in urban areas. In this context, we show that mathematical morphology can enhance methodologies for the processing and analysis of urban remote sensing data. In this paper, we first present a self-adapting method of segmentation of VHRS images. The methodology exploits adjacency graphs and its corresponding morphological processing to describe the scene. Results of this segmentation methodology allows us to obtain suitable and significant connected components. In a second step, the interpretation of the segmented images is realized by means of features extraction according to intrinsic and extrinsic information associated to each region. Geometric and morphological features are first extracted using very simple morphological operations. The image is then described in terms of spatial relations between entities. The proposed methodology is applied to VHRS images interpretation in order to characterize urban areas. We illustrate the usefullness of mathematical morphology tools with PLEIADES HR satellite images.