SAR Image Segmentation by Efficient Fuzzy C-Means Framework with Adaptive Generalized Likelihood Ratio Nonlocal Spatial Information Embedded

The existence of multiplicative noise in synthetic aperture radar (SAR) images makes SAR segmentation by fuzzy c-means (FCM) a challenging task. To cope with speckle noise, we first propose an unsupervised FCM embedding log-transformed Bayesian non-local spatial information (LBNL_FCM). This is measured modified similarity metric which derived applying the distribution to theory. After, construct patches as continued product corresponding pixel generalized likelihood ratio (GLR) avoid undesirable characteristics metric. An alternative framework named GLR_FCM then proposed. In both frameworks, adaptive factor based on local intensity entropy employed balance original and information. Additionally, membership degree smoothing majority voting idea are integrated supplementary optimize segmentation. Concerning experiments simulated images, frameworks can achieve accuracy over 97%. On real work well homogeneous terms region consistency edge preservation.