Skip to Main Content
This work presents a hybrid (geometry- & image-based) technique suitable for providing interactive walkthroughs of large, complex outdoor scenes. Motion is restricted along a smooth predefined path and the input to the system is a sparse set of stereoscopic views at certain points (key-positions) along that path (one view per position). An approximate local 3D model is constructed from each view, capable of capturing photometric and geometric properties of the scene only locally. Then during the rendering process, a continuous morphing (both photometric & geometric) takes place between successive local 3D models, using what we call a "morphable 3D-model". The morphing proceeds in a physically-valid way. For this reason, a wide-baseline image matching technique is proposed, handling cases where the wide baseline between the two images is mainly due to a looming of the camera. Our system can be extended in the event of multiple stereoscopic views (and therefore multiple local models) per key-position of the path (related by a camera rotation). In that case one local 3D-mosaic (per key-position) is constructed comprising all local 3D models therein and a "morphable 3D-mosaic" is used during the rendering process. A partial-differential equation is adopted to handle the problem of geometric consistency of each 3D-mosaic.