Macro-texture mapping from satellite images by morphological granulometries: application to vegetation density mapping in arid and semi-arid areas
The land cover cartography from the optical remote sensing data can be carried out by using the multi-texture classification. In the multi-texture classification, the textural parameters, computed from local or global statistics of the image, are used as the quantitative descriptors to be classified. However, computation of textural parameters from grey-tone values is not relevant for the feature extracted data used in this study. So, textural analysis has to be performed on a binary image. In this paper, we present a method leading to the mapping of the spatial organization, so-called 'macro-texture', of a unique thematic set. The particularity of the proposed method resides in the use of the concept of granulometric analysis of a binary image for defining the macro-texture parameters. The local granulometry density is computed on a window that is centred at each pixel of the binary image. This computation generates grey-tone images that will be classified by using the K-means method. The final map obtained from the classification of the pixels described by macro-textural parameters can be considered as a map of the density of the set under study. The method has been tested and evaluated on Satellite pour l'Observation de la Terre (SPOT) and Landsat Thematic Mapper (TM) data. The data cover the two arid regions characterized by different stages of ligneous covers.