基于Grab Cut和二维熵的SAR图像目标分割方法

A segmentation method for SAR image based on Grab Cut and 2D maximum-entropy

传统的Grab cut算法需要人工交互操作,无法实现SAR图像的自动分割;且SAR图像的斑点噪声干扰容易降低图像的分割质量。针对以上问题,文中以Grab Cut图像分割算法为基础,首先利用FCM算法对SAR图像进行预分割,根据预分割结果标记SAR图像中的前景与背景集合,得到两组较为准确的GMM初始化参数,迭代求得能量函数的最小化,实现SAR图像前景区域与背景区域的自动分割;并结合二维熵算法滤除SAR图像中的阴影,分割出目标区域。实验表明,利用该方法可自动分割出SAR图像中的目标,且分割质量良好。

The traditional Grab Cut algorithm often requires artificial interaction and the segmentation quality of SAR image is lower due to the speckle noise.To solve the above problems,this paper initializes the clusters of the SAR image by FCM algorithm to mark the foreground and background set which gets two sets of accurate parameters of the GMM,minimizes the energy function by the iterative method based on the Grab Cut,and extracts the target and shadow area without any interaction.Furthermore,the target segmentation of SAR image is implemented by 2D maximum-entropy algorithm to clearly divide target and shadow.Experimental result shows that the proposed method can extract the SAR image target automatically and has a good quality.