- An unsupervised anomaly detection algorithm for synthetic aperture radar (SAR) images, making use of polarized data, is developed. The processing contains several stages, including calibration of the images, extraction of information parameters and speckle filtering, detection of candidate pixels and application of a constant false alarm rate (CFAR) morphology operator. The developed algorithm is independent of the anomaly's radar cross section (RCS); it depends only on the physical structure of the observed objects. The proposed processing is non-iterative, adaptive and semi-automatic. Performance evaluation shows improved performance of the algorithm over the common alternatives.