Fuzzy samples retrieval: A method of SAR image retrieval in urban areas

As the retrieval entrance, description of retrieval objects plays an important role in CBSIR (Content-Based SAR (Synthetic Aperture Radar) Image Retrieval) systems. Most of the CBSIR systems take one Region of Interest (ROI) from the original image, which contains the retrieval object, as the retrieval entrance. Besides, freehand sketch retrieval methods are used extensively in CBOIR (Content-Based Optical Image Retrieval). But when the retrieval objects are fuzzy in SAR image, methods mentioned above are no longer applied because speckle noises make the description of retrieval objects unfeasible. In this paper, a method is proposed to describe the retrieval objects in SAR image, which is called Fuzzy Samples Retrieval (FSR). In the experiments FSR is compared with the retrieval based on ROI of original image and freehand sketch sample, the results show that FSR has a better performance in CBSIR systems.