A NOVEL SALIENCY-BASED METHOD FOR SHIP DETECTION IN SAR IMAGE

This paper presents a hierarchical saliency detector for ship detection in synthetic aperture radar (SAR) imagery. First, the nonlinear anisotropic diffusive process has been adopted to eliminate clutter, while preserving the target edge feature in SAR image. Second, each pixel in the filtered image is assigned to its corresponding super-pixel region via an adaptation of optimization technique. Third, Gamma manifold for feature representation has been presented for the modeling of intensity of all super-pixels in SAR imagery. Finally, a threshold segmentation method is used to realize ship detection. Experimental results accomplished over real SAR images demonstrate that the proposed detection method can achieve a good performance.