Conventional dark channel prior (DCP) methods can improve the visibility of scattering-degraded images, but their restoration performance remains limited under non-ideal scattering conditions. In particular, errors in background light estimation, transmission estimation, and model approximation may leave residual background scattering in the restored image, thereby reducing the contrast between the target and the background and weakening the detectability of low-contrast targets. To address this problem, a polarization-difference compensation method is introduced into the DCP framework to further suppress residual background scattering after restoration. Based on the McCartney scattering model, the DCP restoration process is theoretically analyzed, and the residual background scattering term under non-ideal inversion conditions is explicitly derived. The relationship between this residual term and the estimation errors of transmission and background light is clarified. Furthermore, the polarization responses of the target term and the residual background term are modeled in two orthogonal polarization channels, and a differential compensation coefficient is introduced to compensate for the mismatch between residual background components in the two channels, thereby improving background suppression while preserving the target response. To validate the proposed method, an active laser polarimetric imaging system is constructed in a turbid medium prepared using a diluted Intralipid solution. Experiments are performed on a representative target-background configuration in which an aluminum target and a compact disc background exhibit similar reflection intensities but different depolarization responses. The experimental results show that DCP restoration improves the overall visibility of the degraded image, while the proposed polarization-difference compensation further suppresses the residual background response and enhances the target structure. Compared with the DCP result, the compensated result increases the mean target response from 0.492 to 0.748 and reduces the mean background response from 0.409 to 0.112, leading to a more distinct separation between the target and the background. Under different optical-depth conditions, the normalized contrast values remain in the range of 0.799-0.867, indicating that the method maintains effective target-background separation at different levels of scattering degradation. These results demonstrate that the proposed method can jointly exploit intensity and polarization information to compensate for residual background scattering after DCP restoration, providing a physically interpretable approach for target enhancement and backgroundscattering suppression in active imaging through turbid media.