|
Abstract:
I will be discussing some recent work on how to infer probabilistic occupancy representations from a set of N noisy photographs. Based on formal probabilistic definitions of visibility, occupancy, emptiness, and photo-consistency, our development yields a formulation of the Photo Hull Distribution, the tightest probabilistic bound on the shape of the true scene that can be inferred from the photos. I will show how to (1) express this distribution in terms of image measurements, (2) represent it compactly by assigning an occupancy probability to each point in space, and (3) design a stochastic reconstruction algorithm that draws fair samples (i.e., 3D photo hulls) from it. I will also show some experimental results on complex scenes.
This is joint work with Rahul Bhotika (U. Rochester) and David Fleet (Xerox PARC)