A sea-land segmentation scheme based on statistical model of sea

Sea-land segmentation is a key step for target detection. Due to the complex texture and uneven gray value of the land in optical remote sensing image, traditional sea-land segmentation algorithms often recognize land as sea incorrectly. A new segmentation scheme is presented in this paper to solve this problem. This scheme determines the threshold according to the adaptively established statistical model of the sea area, and removes the incorrectly classified land according to the difference of the variance in the statistical model between land and sea. Experimental results show our segmentation scheme has small computation complexity, and it has better performance and higher robustness compared to the traditional algorithms.