Use of Multi-Temporal SAR Non-Local Mean Filtering Operations for Change Detection Analyses

Abstract

The exploitation of sequences of multi-temporal synthetic aperture radar (SAR) images for change detection analyses has become a common practice for the analysis of changes that occurred in different ecosystems. As the first step of any Change Detection approach is the reduction of the speckle effects in every single SAR image. Local and non-local filters have been properly designed to reduce the noise effects and make more efficient the retrieval of changed features. In this work, a joint space-time non-local mean filter, relying on the mutual exploitation of similarities in time and space, is applied to the Kerala region, India, to determine the areas and the extent of a large flood that hit the region in 2018. The methods can be extended for the analysis of different phenomena, such as landslides, coastal flooding, crop monitoring changes depending on the resolution of available SAR images and the number of available SAR scenes that are compared to one another. In this work, some preliminarily results with sequences of Sentinel-1 SAR images are shown.