Computer Engineering and Applications ›› 2021, Vol. 57 ›› Issue (22): 232-240.DOI: 10.3778/j.issn.1002-8331.2007-0260

• Graphics and Image Processing • Previous Articles     Next Articles

Adaptive Region Segmentation of SAR Image Based on Edge Detection

QI Xiaoxiang, LI Min, ZHU Ying, SONG Yu, DU Weidong   

  1. 1.College of Operational Support, The Rocket Force University of Engineering, Xi’an 710025, China
    2.Unit 65367 of the PLA
    3.College of Information and Communication, National University of Defense Technology, Xi’an 710106, China
    4.College of Cryptographic Engineering, Engineering University of PAP, Xi’an 710086, China
  • Online:2021-11-15 Published:2021-11-16



  1. 1.火箭军工程大学 作战保障学院,西安 710025
    3.国防科技大学 信息通信学院,西安 710106
    4.武警工程大学 密码工程学院,西安 710086


Speckle noise degrades the quality of SAR(Synthetic Aperture Radar) image, which makes it difficult to recognize the target from background. Traditional segmentation methods are mostly noise sensitive, lack of detail and over segmentation in SAR images. Therefore, this paper proposes an adaptive region segmentation method for SAR image. Firstly, the cascade filter is constructed by bilateral filtering and Gaussian filtering to reduce the noise in SAR image and preserve the edge information. Then a threshold estimation model based on texture complexity is established for automatic edge detection. Finally, a region growing segmentation method based on edge features is proposed, which solves the contradiction between excessive growing and over-segmentation when utilizing traditional region growing segmentation. This method realizes adaptive region segmentation of single polarized SAR image, comprehensively utilizing the two-dimensional entropy, greyscale values of edge pixels and region gray information. The experimental results verify that this method has stronger ability of edge preservation and noise suppression, more accurate detail detection than others, and overcomes over-segmentation of SAR images partially.

Key words: image segmentation, adaptive threshold, SAR image, edge detection, region growing


受相干斑噪声影响,合成孔径雷达(Synthetic Aperture Radar,SAR)图像成像质量低,目标判读困难。针对传统方法对SAR图像分割存在噪声敏感、细节缺失、过度分割等问题,提出一种基于边缘检测的SAR图像自适应区域分割方法。首先引入双边滤波构建级联滤波器,对SAR图像进行保边抑噪;然后建立基于纹理复杂度的阈值估算模型,实现阈值自适应目标SAR图像边缘检测;最后提出基于边缘特征的自适应区域生长分割方法,较好解决了传统区域生长算法对SAR图像分割时出现的过度生长和过度分割之间的矛盾问题。该方法综合利用了SAR图像二维熵、边缘灰度信息、区域灰度信息,实现了对单极化目标SAR图像的自动分割。实验表明,相较于其他传统分割方法,该方法保边抑噪能力更强,目标细节检测更准确,较好解决了SAR图像过分割问题。

关键词: 目标分割, 阈值自适应, SAR图像, 边缘检测, 区域生长