Perturbation-based significance estimation for multi-temporal and/or multi-polarized SAR descriptors

Abstract:
The reliability of remotely derived information is essential when decisions have to been taken. The direct comparison of remote sensing products to ground truth information often is impaired by various factors. One of them is the system inherent accuracy of the measurement itself. This paper will introduce a novel perturbation-based noise model in order to estimate the reliability of normalized intensity differences often referred to as Kennaugh elements. Beside the theoretical concern of this approach, many interesting applications are feasible such as the automated detection of man-made objects or temporal changes in a landscape.