Generalized Likelihood Ratio Tests for Linear Structure Detection in SAR Images

Abstract:
The detection of linear structures in Synthetic Aperture Radar images is often used as a first step for further processing such as the extraction of road and river networks. In this paper, we propose a new method based on the Generalized Likelihood Ratio Test (GLRT) framework to evaluate at each pixel the likelihood of the presence of a linear structure. Results are presented on Sentinel-1 images and compared with a state-of-the-art method, also derived from the GLRT framework but with a simpler model of the lines. In our experiments, our method produces far fewer false positives than the reference method.