基于参数化模型的水平集SAR图像多区域分割方法

摘要

提出了一种基于参数化模型的水平集合成孔径雷达(SAR)图像多区域分割方法。该方法采用改进的Edgeworth展开式自适应地对SAR图像统计信息进行拟合。由于无需预先估计SAR图像待分割区域的概率密度函数，因此该方法更适用于多区域分割。该方法根据分割区域数量，将改进的Edgeworth展开式嵌入到对应个数的能量泛函模型中，并给出水平集方法求解过程及数值实现方案，最终实现图像多区域分割。实验结果表明，同其他水平集方法相比，该方法能获得更高的分割精度，更适用于多区域分割。%In this paper, a multi-region segmentation method for synthetic aperture radar (SAR) images based on a parametric model is proposed. The modified Edgeworth expansion series method is employed to fit the statistical information of the SAR image adaptively. Since the estimation of the probability density function of the SAR image is not required, it is more suited for the multi-region segmentation. Based on the number of the segmentation, the modified Edgeworth expansion series is introduced to the corresponding energy functional model and then the solution based on the level set and its numerical method are developed, thus attaining the segmentation. The experimental results indicate that the higher accuracy of the segmentation is achieved by the proposed method compared with those by the other methods, and the proposed method is more suitable for the multi-region segmentation.