Visual hulls for large outdoor image collections
Goal
The goal of this project is to detect the sky in a collection of images that capture the same scene and for which the 3D camera position is assumed to be known. Using this, constraints on the structure of the scene can be formulated (known as the visual hull). A possible extention would be to constrain the scene structure based on the 3D point clouds.
Methods
statistical model of the sky appearance
visual hull computation
Type of work
50% C/C++ Implementation, 50% Theory
Contact
Christoph Strecha BC 308
|
|
|
|
|
|
| two views of the 3D point cloud, computed from the images |
two views of the camera positions of all input images |
all input images |