Multiple-View Scenes: Reconstruction and Virtual Views
The problem of generating a virtual view of a scene, i.e. a view from a point where there is not a physical camera to capture the scene, has received recently a lot of attention from the computer vision community. This is probably due to the increase of the computational power of computers, which allows to deal with multiple view systems (systems composed of multiple cameras) efficiently. In this document, an introduction to virtual view generation techniques is presented. In a first part, geometric constraints of multiple view systems are presented. This geometric constraints allow to reconstruct the 3D information of the observed scene, and therefore they allow to generate virtual views from everywhere (although problems with occlusions will arise). In the second part of the document, the state-of-the-art on Image Based Rendering (IBR) techniques is presented. IBR techniques allow to generate virtual views from some constrained regions of the space without requiring a complete 3D reconstruction of the scene. To finish, some concussions are given.