Ship detection with Synthetic Aperture Radar (SAR) is a major topic for the security and monitoring of maritime areas. One of the advantages of using SAR lays in its capability to acquire useful images with any-weather conditions and at night time. Specifically, this paper proposes a new methodology exploiting polarimetric acquisitions (dual- and quad-polarimetric). The methodology adopted for the detector algorithm was introduced by the author and performs a perturbation analysis in space of polarimetric targets checking for coherence between the target to detect and its perturbed version on the data. In the present work, this methodology is optimized for detection of marine features. In the end, the algorithm can be considered to be a negative (notch) filter focused on sea. Consequently, all the features which have a polarimetric behavior different from the sea are detected (i.e., ships, icebergs, buoys, etc). Moreover, a dual polarimetric version of the detector is designed, to be exploited in the circumstances where quad polarimetric data cannot be acquired. The detector was tested with TerraSAR-X quad polarimetric data showing significant agreement with the available ground truth. Moreover, the theoretical performances of the detector are tested with Monte Carlo simulations in order to extract the probabilities of detection and false alarm. An important result is that the detector is, up to some extend, independent of the sea conditions.