This paper deals with an algorithm to reconstruct lunar surface based on a single image or digital elevation model (DEM). To design and test Autonomous precision Landing and Hazard detection and Avoidance Technology (ALHAT), real-time lunar surface simulator is essential. The simulator shall include reconstructed 3D images of lunar surface. Measured data of camera, altimeter and LIDAR can be implemented from the images. Usually, 3D lunar map used in simulator are constructed based on DEM. Since the resolution of provided DEM is unsatisfactory, high-resolution images obtained from vision sensor can be used to improve the resolution of DEM. For 3D reconstruction using 2D camera images, shape from shading (SFS) technique is applied assuming Lambertian reflectance model, following Shah's approach. In this paper, additional information of relative position between lunar and solar is estimated for higher accuracy.