Detection in SAR images based on multi-dimensional generalized low rank model

In this paper, a novel method is proposed for ship detection in polarimetric SAR images. This method employs the multi-dimensional generalized low rank model (GLRM) to model the polarimetric SAR image used for ship detection. An induced multi-dimensional robust principle component analysis (RPCA) framework with a developed fast alternating direction method (ADM) is used to decompose a polarimetric image as a sum of a low rank component, a noise component and a sparse component associated with ships. For comparison, the Polarimetric Whitening Filter (PWF) and Generalized Optimization of Polarimetric Contrast Enhancement (GOPCE) are also used. Experiments results with the RADARSAT-2 C-band image over Singapore Strait show that the proposed method can correctly detect the ships from sea clutter, demonstrating a robust detection performance.