A robust statistic‐aided edge detector for SAR images based on RUSTICO

Abstract

Speckle noise exists inherently due to the special imaging mechanism of synthetic aperture radar (SAR), which is a challenge for edge extraction. The traditional methods failed to make full use of statistical characteristics of SAR images to detect the edges, which results in the loss of weak edges. To address this issue, combined with the robust inhibition-augmented curvilinear operator, this letter proposes a novel edge detector based on the ratio of local statistics (ROLSS) for SAR images. In this letter, the proposed detector achieves the balance between edge resolution and speckle suppression through the proposed intensity calculation method. By taking advantage of image statistic characteristics, the proposed ROLSS edge detector can draw complete continuous edges of many shapes areas and textures without any post-processing, such as edge thinning, extension, and so on. Experiment results show that the proposed detector yields more accurate edge maps with fewer false and missed pixels, and it holds robustness both in the speckle-polluted simulated image and the real X-band airborne SAR image.