AUTOMATED 3D SCENE SCANNING FROM MOTION IMAGES

摘要

A system to automatically generate a fully-textured 3D model of an object from motion images is disclosed. The system tracks salient features in the motion images based on detected salient features using a salient feature operator. A features tracking map is used to construct feature blocks comprising the tracked salient features, each of the feature blocks is then provided as input to a camera motion estimation process that is controlled to provide solutions for a perspective camera motion estimation process, dense points to be used for generating a mesh model are extracted. Finally the mesh model is textured with respect to the motion images so that a fully-textured 3D model is produced.