Geological lineament and shoreline detection in SAR images

In this paper, an algorithm for unsupervised detection of linear structures, in particular, the geological lineament and shoreline, as seen in Synthetic Aperture Radar (SAR) satellite images is proposed. Methodologies to extract linear features consist of three stages. First, the refined local statistics filter is used to remove the speckle noise on the raw image. The likelihood ratio edge detector and local non-maximal suppression are performed on the result images for edge enhancements and detection. In the second stage, mathematical morphology techniques are used to reconnect the fragmented lines and remove the uninterested patterns in the binary image which is produced by thresholding the enhanced image after edge detection. Spatial statistics including spatial mean and spreading coefficient of each cluster (object) in the binary image are calculated in order to simply classify their shapes. The clusters with large spreading coefficient will be remained and others will be removed. Finally, in the third stage, the edge features detected will be described by chain code method and polygonal approximation. The last stage contains many substeps such as edge thinning and curve pruning.