I am working on a project of scene registration where photos taken by the camera on a smartphone are used to build a 3D point cloud of scene. I have an image, the full recon