Superpixel segmentation has been widely used in the field of image segmentation because of its excellent performance. Simple linear iterative clustering (SLIC) shows excellent performance in optical images and has also been widely used in polarimetric synthetic aperture radar (PolSAR) images. However, the initialization method in SLIC cannot accurately locate class centers, which requires many iterations to correct the errors. This paper proposes a new preprocessing method for SLIC. The preprocessing method uses gradient information of the PolSAR images which is calculated by the CFAR edge detector. Experiments using AirSAR PolSAR data show that the preprocessing algorithm can reduce the number of iterations and improve the segmentation accuracy. The segmentation result is more consistent with the image boundaries in comparison with several other classic SLIC superpixel-based segmentation algorithms.