SAR Sea Ice Image Segmentation Using Watershed with Intensity-Based Region Merging

A new approach for Synthetic Aperture Radar (SAR) based sea ice image segmentation for the retrieval of floe size distribution (FSD) is proposed. This method consists of three stages. The first stage involves the pre-processing of the SAR image to reduce the speckle noise in the image by median filtering. The second stage involves an initial segmentation of the image using the Watershed transform. The third stage involves a region merging process based on the difference function of the mean intensity of adjacent regions. Adjacent regions are defined by the region adjacency graph based on a minimum distance of 1 pixel between two adjacent regions. A threshold value is usually set for the merging process, if the difference in mean intensity of adjacent regions is less than the threshold, the adjacent regions will be merged. Experimental results have shown the efficacy of the proposed method in effective segmentation of SAR ice images.