3D Construction From a Sequence of Images

Abstract Building 3D models for the real world from images has attired a lot the intention of researchers in the last years, because this does not require expensive devices, just cameras for the acquisition of images, and this can be viable when it is not possible to use active sensors such as laser scanner. In this paper, we present the full pipeline of building a 3D model from multiple images using a mobile camera. Our algorithm requires a sequence of images taken by an intrinsically calibrated camera, then we estimate the camera positions from the images and a 3D point cloud of the captured scene. From the point cloud, we construct a mesh model then we finish our work by mapping a texture for the mesh model to generate a textured 3D model.