A multi-source remote sensing image matching method using directional phase feature is proposed to solve the problem of matching multi-source remote sensing images with nonlinear radiometric differences. Firstly, feature points of reference image are extracted uniformly. And then phase congruency energy images in multiple directions are calculated using Log-Gabor filters. Dense character descriptions of feature points of the reference image are built. Finally, correspondences are obtained by sliding matching window and Taylor series expansion is used to fit to the sub pixel accuracy. Experiments on three groups of real heterogeneous remote sensing images show that the proposed method can achieve stable and reliable matching results on optical, infrared, multi-spectral and SAR (synthetic aperture radar)images.