Overview of 3D Motion estimation and reconstruction from video streams (presented with Jonathan Balzer, UCLA)

Stefano Soatto
University of California, Los Angeles (UCLA)

We survey the state of the art of computer vision based motion estimation and 3D scene reconstruction for the purpose of visualization (data augmentation, augmented reality) sensor fusion (registration with different modalities, such as OCT), trajectory planning. We describe 2D motion estimation, camera rig calibration and sparse 3D reconstruction of landmarks, moving on to dense 3D reconstruction both passive (using imaging data) and active (using active illuminators or deflectometry).


Back to Long Programs