Single view reconstruction based on marker and geometric constraints

This paper proposes an efficient method for a single image reconstruction based on marker and geometric constraints. Because most of the man-made buildings look like the right prism, we just need to concentrate on the work that how to perfectly reconstruct the right prism. Here we use a calibrated camera and don't concern the lens distortion. In our system, first we build the world coordinate system based on the marker. Second, we get the projection matrix between the scene and the image. Finally, the 3D points can be obtained in three steps: reconstructing the visible points, optimizing the model and recovering the invisible points. Extensive experiments on the cube and the pentagonal prism demonstrate our approach.