In the conventional superresolution, multiple targets are usually separated based on the differences in their locations or Doppler frequencies. The polarization scattering matrix (PSM), as a unique “label” of a target, is very beneficial to radar superresolution, but it is usually neglected. In this paper, a novel scheme of polarimetric superresolution (PSR) is proposed for multitarget detection and localization. We divide the transmit pulse into four parts and modulate the polarization states of each subpulse to derive the sampling in the polarization domain. In the framework of Bayesian compressive sensing (BCS), the estimations of target parameters, including the range, angle and PSM, are derived. However, the off-grid problem results in a number of false targets. To eliminate the false targets, we specially design a true target extractor based on the estimated PSM of each target. With the PSR scheme, each target can be successfully detected and precisely localized in a single pulse. Simulation results demonstrate the effectiveness of the proposed scheme.