An iterative matching method based on choosing optimal mask and representing deformation by the quadric function is proposed to solve the matching problem of 2C level lunar images by Chang′E 1 (CE1) lunar probe satellite. The simulation results show that it has high precision and stability which self adapts well to the relief terrain′s imaging deformation. A precise matching is applied to the normal view, forward view and backward view of the CE1 image by combining the proposed method and the standard correlation method. Affine model approximating the imaging of line scanner camera is analyzed and applied to the same orbit′s three-view reconstruction. The adjacent orbit′s lunar terrain registration is realized by RANSAC estimation of three-dimensional (3D) coordinate transformation parameters led by matching feature points of adjacent normal views. The experiment results indicate that the shape and terrain of lunar surface can be better understood from the reconstructed results.