Spatial filtering strategies on deforestation detection using SAR image textures

Abstract

For SAR images, the inherent speckles have to be suppressed first by employing spatial filtering. However, the spatial filtering is sure to smooth images and lose the textural information. The compromise has to be made between the speckle suppress and the texture extraction in order for a better deforestation detection. In this paper, the focus is to study effects of filtering strategies on the textural information extraction and assess the deforestation detection performance. Adaptive spatial filters are employed in estimating the texture information and constructing the textural change measures. The effects of different filtering strategies on image texture and detection are analyzed using L-band SAR data and the optimized estimation are proposed. The study recommends that the best detection performance appears when the texture map is estimated using a mean filter with a large window size and then the change map is constructed at the single pixel level.