合成孔径雷达（Synthetic Aperture Radar, SAR）是一种主动式成像传感器，可全天时、全天候工作，在国民生活和国防军事等方面有着极大应用潜力。但由于相干成像使图像存在相干斑，影响图像质量，此外舰船结构和成像背景存在特殊性，导致传统的SAR图像分割算法无法进行精准分割。针对以上问题，本文围绕SAR舰船图像相干斑抑制和分割方法展开研究，具体研究内容如下：

针对SAR舰船图像相干斑引起舰船目标灰度特征无法清晰显示在海洋背景上的问题，提出了一种基于加强Lee滤波结合引导滤波的SAR舰船图像相干斑抑制方法（Enhanced Lee Filter with Guided Filtering，ELeeGF）。利用改进的加强Lee滤波对SAR图像切片进行预滤波，去掉切片中大部分相干斑噪声，将结果作为引导图对原图像进行引导滤波，可以有效地消除SAR图像的大多数相干斑，从而既能够获得平滑的图像，又能够保留舰船的轮廓细节信息。经过对比实验表明，本文所提算法等效视数平均增加85.36%，相干斑指数平均降低28.14%，不仅能够很好地对SAR舰船图像进行相干斑抑制，还能够有效地保留图像中的纹理、结构等重要信息。

针对SAR成像时的方位模糊、距离扩散产生的“旁瓣”、“十字叉”、拖影等问题，提出了一种基于模糊C均值聚类（ Fuzzy C-means，FCM）最小距离直线拟合的Radon变换SAR舰船图像分割方法（Radon Transform Segmentation Based on Fuzzy C-means Clustering Minimum Distance Straight Line Litting， FCMIDLF-Radon）。利用改进的模糊C均值聚类将舰船目标与海洋背景进行初始分割，对分割结果进行最小距离直线拟合，拟合的直线作为舰船主体的初始主轴，倾斜角作为舰船目标的初始方位角，再对图像进行Radon变换分割出目标，最后对Radon变换后的图像进行形态学优化。本方法解决了初始方位角确定偏差的问题，仿真结果表明，所提方法在域间差异性、域内一致性有明显提升，形状复杂度平均降低28.23%，更符合舰船的几何形态。

Synthetic Aperture Radar ( SAR ) is an active imaging sensor, which can work all-day and all-weather, and has great application potential in national life, national defense and military. However, due to coherent imaging, the image has speckle, which affects the image quality. In addition, the ship structure and imaging background are special, which makes the traditional SAR image segmentation algorithm unable to segment accurately. In view of the above problems, this paper focuses on the speckle suppression and segmentation methods of SAR ship images. The specific research contents are as follows :

Aiming at the problem that the gray feature of ship target cannot be clearly displayed on the ocean background caused by the speckle of SAR ship image, an Enhanced Lee Filter with Guided Filtering ( ELeeGF ) is proposed. The improved enhanced Lee filter is used to pre-filter the SAR image slice, and most of the speckle noise in the slice is removed. The result is used as a guide map to guide the filtering of the original image, which can effectively eliminate most of the speckle of the SAR image, so as to obtain a smooth image and retain the contour details of the ship. The comparative experiments show that the equivalent number of views of the proposed algorithm increases by 85.36 % on average, and the speckle index decreases by 28.14 % on average. It can not only suppress the speckle of SAR ship images well, but also effectively retain important information such as texture and structure in the image.

Aiming at the problems of azimuth ambiguity, ' side lobe ', ' cross ' and smear caused by range spread in SAR imaging. A Radon Transform Segmentation Based on Fuzzy C-means Clustering Minimum Distance Straight Line Litting ( FCMIDLF-Radon ) is proposed. Firstly, the improved fuzzy C-means clustering is used to segment the ship target and the ocean background, and the minimum distance straight line is fitted to the segmentation result. The fitted straight line is used as the initial spindle of the ship body, and the inclination angle is used as the initial azimuth angle of the ship target. Then the image is segmented by Radon transform. Finally, the image after Radon transform is morphologically optimized. This method solves the problem of initial azimuth determination deviation. The simulation results show that the proposed method has obvious improvement in inter-domain difference and intra-domain consistency, and the shape complexity is reduced by 28.23 % on average, which is more in line with the geometric shape of the ship.