Small target detection under the sea using multi-scale spectral residual and maximum symmetric surround

Detection of small salient targets under the sea is useful for maritime visual surveillance, maritime search and rescue, and ship collision avoidance. In this paper, we propose a small target detection method based on visual saliency map, which is generated by the multi-scale spectral residual and maximum symmetric surround methods. First, the multi-scale spectral residual method is used to detect the targets of different sizes. Then the maximum symmetric surround method is employed to improve the resolution and suppress background for each scale. The final saliency map is obtained by finding the best scale using the 2D entropy. We compare our approach to three salient region detection methods with ground truth and a salient target segmentation application. The experimental results show that the proposed method outperforms the other three methods in both qualitative and quantitative terms.