The interference of specular reflection light such as large-area flare and light spot on target surface during imaging process directly hinders the vision performance like detection and matching.A specular reflection separation algorithm based on polarization imaging is proposed, which suppresses the impact of specular reflection on target surface.By analyzing the polarization imaging mechanism of dichromatic reflection model, and using the polarization characteristics difference between diffuse reflection and specular reflection, the inversion mechanism of the intensity of diffuse reflection based on Stokes parameters is proposed.Then, the constraint of spatial CbCr-chromaticity vector is constructed, and the physical isolation of specular light in CbCr space is realized.The traversal least squares method is applied to obtain the reflection parameters.Thus the specular component is effectively removed, and the inherent local details on target surface are restored.Experimental results show that the proposed algorithm can achieve accurate specular reflection removal.Meanwhile, the original color, structure and texture features are highly preserved.The algorithm is scene-adaptively robust with high recovery capability.