基于优化活动轮廓模型的SAR影像海陆分割方法研究*

SAR（合成孔径雷达）影像具有很强的乘性斑噪，给图像分割带来了困难。本文利用Gamma分布拟合SAR影像，并将其用于构造基于区域信息的能量泛函，提出了一种基于活动轮廓模型的SAR影像海陆自动分割方法。该方法在能量泛函中同时融合了边缘信息和区域信息，既有利于边界精确定位又有利于降低乘性斑噪的影响，利用活动轮廓演化模型，通过变分水平集方法推动活动轮廓曲线向海岸线演化，在最小化特定的能量泛函的约束下，使活动轮廓与海岸线重合，达到影像分割的目的。同时针对该模型提出了优化方法提高其计算效率，使本文提出的分割算法更加实用。

Sea-land segmentation based on advanced active contour model for SAR imagery

The multiplicative speckle noises brought great difficulty to SAR(Synthetic Aperture Radar) image segmentation. This paper adopted Gamma distribution to fit the image and used it in the construction of the region information based energy function. Thus an automatic active contour model based sea-land segmentation method was proposed for SAR imagery. This method combined both edge information and regional information in the energy function which was propitious to both border accurate positioning and decreasing the multiplicative speckle noise effect. The model forced the active contour to the real coastline by variation level set. Under the restriction of minimizing the constructed energy function, the method obtained the final segmentation result when the active contour was overlapped on the coastline. In addition, an optimizing method was proposed to improve its computation efficiency, which made our segmentation method more applicable.