Waterline Extraction Based on Superpixels and Region Merging for SAR Images

Waterline extraction is of importance for safe navigation and environment protection. However, the existing pixel-based method failed to get accurate edges due to the similar features between land and interference. This paper proposes a minimum ratio with the mean and standard variation as similarity measure, and presents an improved region merging criterion which based on superpixels for waterline extraction. Utilizing the Envisat and Radarsat images, experiments show the proposed method can effectively extract the waterline and demonstrate better performance in contrast with some existing pixel-based algorithms.