SAR图像超像素生成算法抗噪性能研究

Research on anti-noise performance of superpixel segmentation algorithms of SAR image

合成孔径雷达(synthetic aperture radar,SAR)图像受相干斑噪声影响严重,针对SAR图像的超像素生成算法需具有较强的抗噪性能,现有的SAR图像超像素生成算法有很多种,但对其抗噪性能的研究并不多。文章针对上述问题进行研究,基于区域冗余度和区域准确率,提出一种SAR图像超像素生成算法的抗噪性能评价方法,对几种经典的SAR图像超像素生成算法的抗噪性能进行评价。实验采用不同噪声水平的合成SAR图像以及由SIR-C和RADARSAT-2获取的真实SAR图像进行测试。结果表明,与其他算法相比,efficient graph-based segmentation(EG)算法的抗噪性能最优,最适用于SAR图像分割。

Superpixel algorithm for synthetic aperture radar（SAR） image requires strong anti-noise performance due to the speckle noise. Although many superpixel algorithms for SAR image have been proposed, there are few researches on their antbnoise performance. In this paper, based onthe region accuracy and region redundancy, a method to compare the anti-noise performance of several classical superpixel algorithms for SAR image is proposed to find out the best anti-noise algorithm. The artificial SAR images with different noise levels and the real SAR images obtained by SIR-C and RADAR- SAT-2 are used in the experiment, and the experimental results show that the efficient graph-based segrnentation（EG） algorithm is superior to other three algorithms in the anti-noise performance, which is more suitable for SAR image segmentation.