Compact polarimetric (CP) synthetic aperture radar (SAR) has proven its potential in distinguishing oil slicks and look-alikes. Polarimetric information can be retrieved directly from scattering vector or from reconstructed pseudo-Quad-Pol covariance matrix of CP SAR data. In this paper, we analysed features from Circular Transmit and Linear Receive (CTLR) CP SAR data that are derived by taking both of these two methods. K-means clustering followed by accuracy assessment was also implemented for performance evaluation. Through experiments that were conducted based on L-band UAVSAR fully polarimetric data, it was found that optimum extraction methods varied for different features. The histogram analysis and segmentation results also demonstrated the comparable performance of CP SAR features in distinguishing different damping properties within oil slicks. This study proposed a framework of statistically analyzing polarimetric SAR (Pol-SAR) features and provided guidelines for determining optimum feature extraction methods from CP SAR data and for marine oil-spills detection and classification.