ABSTRACT
This chapter examines some commonly used techniques for enhancing “change signals” in bi-temporal satellite images. It focuses on the multivariate alteration detection (MAD) algorithm for visible/infrared imagery and on a change statistic for polarimetric synthetic aperture radar data based on the complex Wishart distribution. The chapter deals with an “inverse” application of change detection, in which unchanged pixels are used for automatic relative radiometric normalization of multi-temporal imagery. Interesting small-scale anthropogenic changes will generally be unrelated to dominating seasonal vegetation changes or stochastic image noise, so it is quite common that such changes will be concentrated in lower-order MAD variates. An additional advantage of the MAD procedure stems from the fact that the calculations involved are invariant under linear and affine transformations of the original image intensities. The MAD transformation can be augmented by subsequent application of the MAF transformation in order to improve the spatial coherence of the MAD variates.
