Abstract
In this contribution we present a gradient-based approach for edge detection in multi-dimensional images and its application to multi-spectral satellite imagery. Central to the problem of multi-spectral edge detection is the question of how to integrate the contrast information contained in the various channels into one meaningful result. We show how to extend in a formal way a state-of-the-art grey-level edge detector to multi-dimensional image data. The integrating approach combines the contrast information coming from different spectral channels in a well-founded way. It takes advantage of the spectral redundancy, allowing at the same time for the integration of uncorrelated information. Based on this approach we describe a detection scheme which can handle an arbitrary number of image components (spectral channels). In several experiments carried out on data from Landsat-TM we have tried to demonstrate the kind of results which can be expected in practical applications.