ETD

• assisted reconstruction
• dense stereo reconstruction
• image-based modeling
• multi-view stereo

This thesis faces the problem of automated acquisition of object and scene models from multiple photographs, a process known as image-based modeling, or 3D photography.

We investigate a new paradigm for the acquisition, making the process interactive and almost real-time, and allowing the quality of the final result to be controllable, in contrast with existing methods. The reduced computational resources needed by this incremental approach allow the task to be affordable on average-end hardware, such laptops.

In addition, we propose a tool which implements this paradigm, showing that notable gains in terms of computational time can be achieved in the dense stereo reconstruction step, still maintaining a quality of the final result that is comparable, or better, than that achievable with existing methods.