Abstract: For the purpose of automatic and effective stereo image matching in the field of digital close-range stereo photogrammetry, a new stereo image matching algorithm based on Digital Parallax Model (DPM) and improved SIFT feature is proposed in this paper. In the new algorithm, considering characteristics that baseline and relation of two cameras is fixed in digital close-range photogrammetry system, digital parallax model of stereo image is constructed firstly, and computation parameters for confirming conjugate epipolar line are work out with known calibration data. Then searching scope of conjugate points on the right image is confirmed under constraints of DPM and epipolar line to improve matching speed and accuracy. Secondly, to increase matching robustness, the computation model of matching similarity measure combined with local structure information and global gray-value information is designed by SIFT feature and gray-value feature. The new algorithm is applied to actual stereo images taken by digital close-range photogrammetric system. Experimental results show that the proposed algorithm can decrease matching error greatly, and enhance the reliability and precision of stereo image matching.