Multiresolution analysis and reconstruction by a morphological pyramid in the remote sensing of terrestrial surfaces

This paper investigates the use of morphological tools for the multiresolution analysis of remote sensing imagery in the field of spatial dynamics on terrestrial surfaces. It presents a multiresolution analysis which exploits the properties of the non-linear operators of mathematical morphology. The aim is to break down the overlapping of elements constituting the organization of landscapes. The methodology behind multiresolution analysis is then presented, based on a morphological pyramid which links a physically significant morphological filter with a dyadic decomposition. The results of the inverse transformation are described by reconstruction of the images. These results enable the photometry stability of the transformation to be verified. An application is described in a real case of observation with satellite images. The interpretation of the results shows the successive information which are available at different stages of analysis. (C) 2001 Elsevier Science B.V. All rights reserved.