Meaningful Objects Segmentation from SAR Images via A Multi-Scale Non-Local Active Contour Model

TL;DR

This paper introduces a multi-scale non-local active contour model for SAR image segmentation that effectively handles speckle noise and backscattering variations, improving accuracy across different sensor types.

Contribution

The paper proposes a novel multi-scale non-local active contour approach that addresses both speckle noise and backscattering variability in SAR image segmentation, enhancing robustness and precision.

Findings

Accurate segmentation of SAR images with heavy speckle noise.

Effective handling of non-local intensity variations.

Applicable to SAR images from various sensors.

Abstract

The segmentation of synthetic aperture radar (SAR) images is a longstanding yet challenging task, not only because of the presence of speckle, but also due to the variations of surface backscattering properties in the images. Tremendous investigations have been made to eliminate the speckle effects for the segmentation of SAR images, while few work devotes to dealing with the variations of backscattering coefficients in the images. In order to overcome both the two difficulties, this paper presents a novel SAR image segmentation method by exploiting a multi-scale active contour model based on the non-local processing principle. More precisely, we first formulize the SAR segmentation problem with an active contour model by integrating the non-local interactions between pairs of patches inside and outside the segmented regions. Secondly, a multi-scale strategy is proposed to speed up the non-local active contour segmentation procedure and to avoid falling into local minimum for achieving more accurate segmentation results. Experimental results on simulated and real SAR images demonstrate the efficiency and feasibility of the proposed method: it can not only achieve precise segmentations for images with heavy speckles and non-local intensity variations, but also can be used for SAR images from different types of sensors.