基于边缘强度映射的SAR图像等效视数非监督估计方法

该文结合边缘强度映射(Edge Strength Map, ESM)对SAR图像等效视数(Equivalent Number of Looks, ENL)，提出一种自适应非监督估计方法。该方法先利用各向异性高斯核平行窗通过比率运算生成ESM，进而在简单的图像分块处理基础上，利用一种有效的非监督估计方法分别估计各分块图像在比率ESM下的区域阈值，并由各分块图像的阈值化ESM操作完成对SAR图像的区域划分，然后对非边缘区域中的所有像素在ESM的两次约束下，通过施加大尺度不规则窗完成局部ENL估计，最后通过直方图统计方法从局部ENL中获得全局ENL估计。实验表明：该文方法可有效克服传统空域估计方法ENL估计值受局域统计方法选择影响大的不足，同时也可避免基于统计模型的估计方法需要相干斑先验信息且对抑斑后SAR图像估计失效的弊端。

Unsupervised Estimation of the Equivalent Number of Looks Based on Edge Strength Map in SAR Images

Combining with Edge Strength Map (ESM), an unsupervised estimation of the Equivalent Number of Looks (ENL) in SAR images is proposed. First, ESM is produced by ratio operation of anisotropic Gaussian kernel parallel windows. On the basis of simple image partition processing, the local ESM thresholds for image blocks are obtained by an effective unsupervised estimation method, and SAR image is divided into homogeneous regions and edge regions by shrinkage ESM. Under twice restrictions based on ESM, each local ENL of all pixels in homogeneous regions is estimated by irregular window with large scale. Then global ENL of SAR image is acquired by histogram statistical method for all local ENL values. Experimental results show that the proposed method can get over serious adverse effects to estimation value by the selection of local statistics methods in estimation methods based on spatial statics, and can avoid the need for apriority information on speckle and being an ineffective estimation for despeckling SAR images in estimation methods based on statistical model.