Interval edge estimation in SAR images

TL;DR

This paper introduces a bootstrap-based method for estimating confidence intervals of edges between textured regions in SAR images, addressing the challenge of speckle noise and improving edge detection reliability.

Contribution

It proposes a novel bootstrap-based approach for interval edge estimation in SAR images, enhancing the assessment of edge detection accuracy under speckle noise.

Findings

Interval estimates are accurate and useful for edge validation.

Interval edge estimation can detect the absence of an edge.

Method validated with Monte Carlo simulations and real data.

Abstract

This paper considers edge interval estimation between two regions of a Synthetic Aperture Radar (SAR) image which differ in texture. This is a difficult task because SAR images are contaminated with speckle noise. Different point estimation strategies under multiplicative noise are discussed in the literature. It is important to assess the quality of such point estimates and to also perform inference under a given confidence level. This can be achieved through interval parameter estimation. To that end, we propose bootstrap-based edge confidence interval. The relative merits of the different inference strategies are compared using Monte Carlo simulation. The results show that interval edge estimation can be used to assess the accuracy of an edge point estimate. They also show that interval estimates can be quite accurate and that they can indicate the absence of an edge. In order to illustrate interval edge estimation, we also analyze a real dataset.