Abstract:

Multiview videos are widely used for 3-D visualization. Multiview images of a scene or an event are commonly captured by people holding different types of cameras. Such content can be used for 3-D visualization applications. In such applications, correspondences are used for 3-D visualization. These correspondences are normally sparsely estimated using feature-based searches. However, a dense correspondence is highly preferred if one concerns high quality of 3-D visualization. Due to the differences in views in terms of sampling density, frame rate, and field of view, a dense detection of correspondences is not straightforward. 

In this project we aim to capture multiview video by a heterogeneous camera array and develop its rectification algorithms. 

The project will have following step:

1. Set up a heterogeneous camera array.

2. Develop synchronization mechanisms for capturing.

3. Geometry and color calibration of the array.

4. Rectification of the heterogeneous camera array.

Keywords:

Camera array, Image analysis, computer vision

Supervisor:

Mehrdad Teratani (mehrdad.teratani@ulb.be), Gauthier Lafruit (gauthier.lafruit@ulb.be)

Director:

Mehrdad Teratani (mehrdad.teratani@ulb.be)