基于图割及均值漂移的合成孔径雷达图像强散射目标分割

针对合成孔径雷达(SAR)图像的特点及标准图割分割算法计算量较大等问题，提出了一种基于图割及均值漂移(Mean Shift)的高效的SAR图像强散射目标分割方法。该方法利用均值漂移算法对SAR图像进行预处理，将原图像表示为基于过分割区域的图结构;然后，以这些过分割图像区域为节点建立区域邻接图，运用图割分割算法得到SAR强散射目标的分割结果。与标准图割算法中以单像素为节点构建邻接图相比，参与图割算法的节点和边的数目减少了两个数量级，计算效率大幅提高。另外，根据SAR图像中目标的强散射特性，自动定义终端节点，减少了人工交互量。实验表明，该方法充分利用均值漂移及图割的优点，能够在背景杂波的干扰下有效地提取SAR强散射目标。

Strong scattering objects segmentation based on graph cut and Mean Shift algorithm from SAR images

Aiming at the characteristics of Synthetic Aperture Radar (SAR) images and the problem of the standard graph cut segmentation algorithm's high computational complexity, a method of strong scattering objects segmentation based on graph cut and Mean Shift algorithm was proposed. Firstly, the image was pre-processed with the Mean Shift algorithm to produce over-segmentation areas. Then, a graph was built with nodes responding to over-segmentation areas, and then the results of SAR strong scattering targets segmentation were obtained by using graph cut algorithm. Compared with nodes responding to pixels in the standard graph cut algorithm, the number of nodes and edges in the graph were reduced by two orders of magnitude and the computational efficiency was significantly improved. Furthermore, according to the strong scattering characteristics of the targets in SAR images, the “object” terminal and the “background” terminal were defined automatically to reduce human interaction. The experiments show that the proposed method combines the advantages of Mean Shift and graph cut effectively, and it can effectively extract SAR strong scattering targets from the background clutter.