Abstract: Although the weighted average ratio （ROEWA） edge detection operator is suitable for SAR images, this operator can not decide the edge direction according to gradient. Since SAR image is affected by multiplicative noise, it is taken the logarithm to change multiplicative noise to additive noise, then Canny operator can be used to calculate the direction of edges. Since land image has a lot of edges and ocean space is flat, land and ocean is different in gray scale, sea and land can be segmented, region growing is also used. Results show that the improved ROEWA operator is efficiency and precision in sea-land segmentation, the robustness is also good.

Key words: SAR image, ROEWA operator, Canny operator, edge detection, sea-land segmentation

摘要： 加权指数平均比率（ROEWA）边缘检测算子是一种较好的适用于SAR 图像的边缘检测算子，但是使用梯度计算的方法不能准确地确定边缘的方向。针对这一问题，由于SAR图像受乘性噪声干扰，对图像取对数，将乘性噪声转化为加性噪声的形式，结合Canny算子计算边缘方向。根据陆地图像边缘丰富，海洋区域平滑的特点，以及陆地和海洋在灰度上的差异，结合区域生长完成海洋和陆地的分割。实验表明将改进的ROEWA算子用于海陆分离，检测效率和精度都比较高，且鲁棒性好。

关键词: SAR图像, ROEWA算子, Canny算子, 边缘检测, 海陆分离