Ship extraction is usually a challenging problem because of different complicated backgrounds, including coarse sea surface, clouds and shadows. A new cloud over ship model is proposed according to the features of intensity and whether the cloud has a shadow or not. Firstly, shape prior is introduced based on shape distance metrics. An energy function is then constructed in which shape energy is computed according to the similarity of the shape information of the ship and its corresponding shape template. Secondly, the energy function is minimized based on graph cut theory. The experiment results show efficiency of ship extraction in cloudy sea backgrounds in optical remote sensing images whether there is a cloud shadow or not.