WECS: A wavelet energy correlation screening based method for unsupervised change detection using remote sensing image series

Abstract

This paper introduces an unsupervised method for detecting spatiotemporal changes in a series of remotely sensed images. Specifically, we employ a fully automatic, data-driven framework that incorporates wavelet approximation, wavelet energy apportionment, and high-dimensional correlation screening of wavelet coefficients. This approach processes sequences of images and produces a mapping of changed and non-changed locations over the analyzed period.